Panagiotis Kolokathis

Panagiotis Kolokathis (Nova Mechanics)

Panagiotis D. Kolokathis, PhD, is a computational scientist who graduated from the Chemical Engineering School of National Technical University of Athens (NTUA). He holds a Master's Degree in Quality Management and Technology and a PhD in the field of material science and engineering. After completing his studies, he worked for 5 years as a postdoctoral researcher at NTUA in collaboration with Fahrenheit GmbH, Fraunhofer ISE, and the University of Leipzig on the  development of next-generation solar cooling adsorbers. He then worked as a R&D plastics’ engineer at Lalizas SA before joining NovaMechanics as a computational scientist. Kolokathis' research interests  focus on density functional theory, statistical mechanics, atomistic and coarse-grained simulations,  transition state theory, computational fluid dynamics, Lagrangian Dynamics, Rietveld refinement,  machine learning, design of experiments, and invention of new scientific algorithms/methodologies.  He has expertise in simulations of Nanocomposite Materials. Some of his notable accomplishments  include identifying and determining the hydrated ALPO-5 structure for the first time, providing an  analytical solution for the Master Equation for calculating the diffusion coefficient inside periodic  materials, and conceiving and developing the KoBra software. He speaks Greek, English, French, and  German.

Materials modelling supporting hazard profiling

Materials such as cellulose are promising bio-based ingredients for industries (e.g., textile, cosmetics, textile, automotive).  The chemical environment that these materials are exposed to (e.g., water, methanol, acetone, benzene, methanol, oxygen, nitrogen, air or vacuum) is significant in investigations of their function, safety and sustainability for specific applications. We consider for investigation the effect of the substitution of functional groups to achieve desired properties. Physics-based modelling and simulation approaches provide a complementary approach to data-driven machine learning approaches for this purpose.

Functionalisation of the surface of materials can change the descriptors and the properties of materials. Our predictive model aims to make an initial screening among candidate functional groups to find the most promising ones for further investigation. A consensus prediction model has been developed based on the predictions of k-nearest neighbours, Random Forest and Random Trees models for the prediction of skin toxicity while builders of nanoparticles have been created as web applications. These builders calculate atomistic descriptors (e.g. surface tension) while they also produce configuration files needed for subsequent atomistic simulations.

We will present here preliminary modelling results obtained on the SSbD4CheM cosmetics case study.