Research

a couple of presentations: 

                                                               (scroll down to see videos!)

----------------------------------------------------------

New methods for the thermodynamics and kinetics of activated processes

In many important processes - like chemical reactions, phase tranitions, transformations of nanostructures, or protein folding - different metastable states interconvert very slowly because of the presence of barriers on the free-energy landscape. 

Molecular dynamics (MD) simulations are very useful but are limited to short time scales (typically less than one nanosecond ab initio and less than a millisecond using empirical force fields). We develop new approaches to sample efficiently such activated processes with MD, exploiting as well as possible the available computer resources.

In techniques like metadynamics and umbrella sampling this result is achieved adding an external bias during the simulation. The bias is then systematically removed from the results of the simulations to recover the equilibrium free energy landscape and kinetics of the original (unbiased) system. The new methods are implemented in freely available software tools, including plumed, metagui, piv-clustering etc. (see Download page).

We recently developed a set of techniques to directly extract free-energy barriers and kinetic rates from a small amount of short, unbiased MD trajectories: a great advantage is that the reaction coordinate (order parameter) can be optimized at the same time, without requiring additional simulation data. The key tools are Langevin equations and a recently proven variational principle that connects the optimal order parameter with the minimal kinetic rate (see Papers page, 2022-2023).

----------------------------------------------------------

 Chemical reaction dynamics in gas phase and in solution

Simulating the dynamics of chemical reactions, especially in explicit models of solutions including hundreds of atoms, is a difficult task: high barriers must be crossed, and the relevant slow degrees of freedom tracking the transformation (reaction coordinates) are hidden in a haystack of molecules diffusing randomly. Departing from the usual approaches, we introduce new formulations of reaction coordinates borrowing concepts from the study of complex networks (graph theory). The main idea is to regard the system as a network of atoms, described by an adjacency matrix whose information is coarse-grained into a space of topological coordinates.

In this way we could develop new methods addressing in a unified way a variety of different reaction mechanisms, and passing seamlessly from gas phase to solutions. The main results are that reaction pathways and products can be discovered without the need to guess them, and that it is possible to avoid the steep learning curve that usually affects free energy calculations.

We discovered interesting new features of systems like amine solutions for the capture of CO2 from industrial flue gases, or prebiotic reaction networks based on formamide and hydrogen cyanide. We aim at systematically understanding the effects of solvent, temperature and pressure on reaction dynamics.

----------------------------------------------------------

Transformation mechanisms of nanostructures and solids

The properties of a material and its potential for technological applications are determined by its behavior at realistic conditions of temperature, pressure, etc. In addition, to design viable preparation routes of new materials we must understand the details of possible transformation pathways at different conditions. MD simulations are a powerful in-silico microscope to explore atomic-scale mechanisms.

We extend the reach of simulations by developing enhanced sampling methods based on graph theory (see also previous section). In this way we could discover new complex transformation pathways of, e.g., graphene-based nanostructures. We recently invented a general method addressing structural phase transitions including crystals, amorphous solids, and liquids. We can now simulate disorder to order transitions like ice nucleation. 

----------------------------------------------------------

Protein dynamics:  complexes, ligand binding, folding

Proteins, the work-horses of all living organisms, are complicated chains made by tens to hundreds of different amino acids, and assuming a range of possible conformations. We employ MD simulations to address a series of challenging and exciting problems, such as: how do proteins fold into a well-defined 3D structure? What is the detailed assembly mechanism of protein complexes? Can we design drugs interfering with complexes, in particular with transient mis-bound configurations, perturbing protein interaction networks in a controlled way? Can we understand amyloid aggregation of misfolded proteins, linked to neurodegenerative diseases? The computed structural, thermodynamic and kinetic properties are compared with experiments to help rationalize them. We introduced methodologies for inducing changes of secondary structure and to reconstruct high-dimensional free energy landscapes and kinetic rate networks that are presently used by a number of laboratories worldwide.