From Bouncing to Making a Splash: Computational Modelling of Impact Across Scales
by Assoc. Prof. Radu Cimpeanu, University of Warwick

Date: Tuesday, 20 September 2022


The canonical framework of drop impact provides excellent opportunities to co-develop experimental, analytical and computational techniques in a rich multi-scale context. The talk will represent a journey across parameter space, as we address beautiful phenomena such as bouncing, coalescence and splashing, with a particular focus on scientific computing aspects and associated numerical methods.

To begin with, consider millimetric drops impacting a deep bath of the same fluid that are generated using a custom syringe pump connected to a vertically-oriented needle. Measurements of the droplet trajectory are compared directly to the predictions of a quasi-potential model, as well as fully resolved unsteady Navier-Stokes direct numerical simulations (DNS). Both theoretical techniques resolve the time-dependent bath interface shape, droplet trajectory, and droplet deformation. In the quasi-potential model (building on recent progress by Galeano-Rios et al., JFM 912, 2021), the droplet and bath shape are decomposed using orthogonal function decompositions leading to a set of coupled damped linear oscillator equations solved using an implicit numerical method. The underdamped dynamics of the drop are directly coupled to the response of the bath through a single-point kinematic match condition, which we demonstrate to be an effective and efficient technique. The hybrid methodology has allowed us to unify and resolve interesting outstanding questions on the rebound dynamics of the multi-fluid system.

We then shift gears towards the much more violent regime of high-speed impact resulting in splashing, where a combination of matched asymptotic expansions grounded in Wagner theory and DNS allow us to produce theoretical predictions for the location and velocity of the ejected liquid jet, as well as its thickness (Cimpeanu and Moore, JFM 856, 2019). While the early-time analytical methodology neglects effects such as surface tension or viscosity (focusing on inertia instead), corrections and adaptations of the technique (Moore et al., JFM 882, 2020) will also be presented and validated against an associated computational framework, bringing us even closer to efficiently providing information of interest for applications such as inkjet printing and pesticide distribution.


Machine Learning for Polymer Design in the Case of Limited Training Datasets

by Dr Sergey V. Lyulin, Head of “Theory and Modelling of Polymer Systems” Laboratory, Institute of Macromolecular Compounds, Russian Academy of Sciences

Date: Tuesday, 7 June 2022


Machine learning (ML) represents a novel theoretical approach in computational material science which yields the modern paradigm of virtual material design based on known experimental data. The main motivation to use ML methods in computer-aided material design is based on the following fact: the total number of small organic molecules is about 1060, while the number of currently known ones does not exceed 108.

Focusing on polyimides (PI), which are high-performance heterocyclic polymers with big variance of chemical structure, we developed graph convolutional neural network (GCNN), being one of the most promising tools for working with big-data, to predict their glass transition temperature (Tg) as an example of the most fundamental polymer property. To train developed GCNN, we propose an original methodology based on using “synthetic” datasets for pre-training and an available, rather small experimental dataset for its fine-tuning. Using known “chemical rules” we combinatorically generated a huge “synthetic”  dataset of PI chemical structures and calculated their temperature of glass transition Tg with the help of Askadskii's QSPR computational scheme.

By using our database, we developed GCNN which allows estimating Tg of PI with mean absolute error (MAE) around 20 degrees that is almost two times lower than in the case of using quantum-chemically calculated dataset for small molecules (QM9). The proposed methodology may be generalized for predicting any other polymer properties.

This work has been financially supported by by the Russian Science Foundation, grant No. 22-13-00066.


In-Silico Studies of Polyelectrolyte Membranes for Flow Batteries and Fuel Cells

by Assoc. Prof. Alexey V. Lyulin, Eindhoven University of Technology, The Netherlands

Date: Tuesday, 31 May 2022


Nafion is a commonly used polyelectrolyte membrane (PEM) in fuel cells and flow batteries. Nanocomposites of Nafion are used to enhance temperature resistance and proton conductivity. The properties of hydrated membranes, and the water influence on Nafion glassy behavior is very important. We first report molecular-dynamics simulations of Nafion and compare the results with the conductivity of PFIA membrane [1].

Finally, we report the modelling of the annealing effects on both structure, dynamics and electric conductivity of the Nafion membranes. We observe [2] strong antiplasticization effect and increase in the glass-transition temperature upon hydration. The hydrophilic channels evolution (see Figure below) upon annealing and associated changes in ion diffusion and electric conductivity will be discussed.

[1]. S. Sengupta, A. V. Lyulin, J. Phys. Chem. B, 122, 6107-6119, 2018.
[2]. A. V. Lyulin S. Sengupta, A. Varghese, P. Komarov and A. Venkatnathan, ACS Appl. Polym. Mater., 2, 5058-5066, 2020


Biomolecules Under a Computational Microscope: Protein Structure-Dynamics-Function Relationships by Unifying Molecular Simulations with CryoEM Experiments

by Dr. Faidon Brotzakis, IESL-FORTH

Date: Tuesday, 17 May 2022


Molecular simulations have been instrumental in identifying the structure-function relationships of biomolecules at the atomic level as well as providing a means for structure-based drug discovery, thereby explaining and guiding experimental findings.

The increase in computational power, new physics and machine-learning-based algorithms are significantly driving the boost in the field and give access to addressing biomolecular phenomena of increasing length and timescales. In this talk, he will discuss examples where, using state-of-the-art integrative structural biology methods that inject Cryo-EM experimental data into the simulation, accurate protein-functional dynamics of the SARS-CoV-2 spike protein at an atomistic level are revealed.

In this way, the following is achieved: a) reveal virus vulnerabilities by identifying cryptic binding sites exposed during the S protein conformational transition related to the recognition to the host cell [1] and b) provide with the molecular motion and energetics of protein-antibody complexes which enables to suggest mutations that increase the spike-antibody affinity [2]. These predictions are validated in further CryoEM experiments. In this talk, he will also highlight Cryo-EM/simulation-based structural dynamics-function relationships of a tau-microtubule complex implicated in AD [3].


1. Brotzakis, Z. F.; Lohr, T.; Vendruscolo, M. Determination of intermediate state structures in the opening pathway of SARS-CoV-2 spike using cryo-electron microscopy. Chem. Sci. 2021, 12, 9168
2. Huo et al. (unpublished) 2022
3. Brotzakis, Z. F., et al. A Structural Ensemble of a Tau-Microtubule Complex Reveals Regulatory Tau Phosphorylation and Acetylation Mechanisms. ACS Cent. Sci. 2021, 7 (12), 1986-1995


Advanced Material Processing with Ultrashort Pulsed Lasers Through a Detailed Description of the Fundamental Multiscale Physical Processes

by Dr. George Tsibidis, IESL-FORTH

Date: Tuesday, 5 April 2022


Over the past decades, the use of ultrashort pulsed laser sources for material processing and associated laser-driven physical phenomena has received considerable attention due to the important technological applications, in particular in industry and medicine. Various types of surface structures generated by laser pulses and, more specifically, the so-called laser-induced periodic surface structures (LIPSS) on solids have been studied extensively. 

Several laser-driven processes, including energy absorption, photo-ionization processes, electron excitation, electron-relaxation mechanisms, phase transitions and/or thermomechanical effects, resolidification, and mass ejection are some complex and highly nonlinear phenomena that need to be described and be coupled into a multiscale physical theoretical model to explain the self-organisation-based LIPSS formation. A thorough understanding of these laser-driven physical phenomena are expected not only to lead to an enhanced control of the laser energy for numerous potential applications but contribute to the fundamental understanding of the underlying physical processes. Hence, it is of paramount importance to develop appropriate modelling schemes that couple together processes on various timescales and provide precise and consistent description of the electron dynamics and lattice response following irradiation of a solid with ultrashort pulses.  

To address the above challenges, Dr. Tsibidis will briefly present his research efforts towards the state-of-the-art in the development of a multiscale physical model to efficiently describe surface modification processes on various types of materials.

Presentation slides


Multiscale Simulations of Polymers at Interfaces

by Prof. Paola Carbone, University of Manchester

Date: Tuesday, 8 March 2022


The prediction of the adsorption and dynamic properties of polymers at solid and soft interfaces is an important technological and biological problem. Solid interfaces are indeed present in all polymer composites (where particles are dispersed into a polymeric matrix with the aim of improving its mechanical and rheological properties) but also relevant for many applications such as for coatings. Polymers also adsorb at liquid interfaces in many industrial processes, such as liquid/liquid extractions, solvent displacement methods, or emulsifications, and also when used for biological applications, such as drug nanocarriers, biocompatibilizers, or protective coatings.

In this talk, the speaker will show how multiscale modelling can help in predicting the adsorption properties of polymers at solid surfaces (specifically carbon black) and soft interfaces (liquid/liquid). The speaker will clarify the thermodynamics of adsorption and how the properties of the interface as well as of the bulk polymer change upon adsorption.


Contact Line Dynamics in Heterogeneous Environments

by Assist. Prof. Nikos Savva, CaSToRC, The Cyprus Institute

Date: Tuesday, 22 February 2022


Moving droplets on surfaces is an ubiquitous phenomenon in the natural world, which also finds application in a broad spectrum of technologies. Yet, their study is inherently complex and many of these applications are being developed based on intuition derived from laboratory observations. This complexity stems from the multiscale nature of the phenomenon, from forces that manifest themselves at the macro-scale, such as gravity and capillarity, to the micro-scale effects close to the droplet front. Although impressive progress in contact line phenomena and the broader field of wetting hydrodynamics has been made in recent decades, there are several open questions that still elude us, including the poorly understood dynamics on surfaces that contain small-scale heterogeneous features.

The talk will give an overview of the challenges in modelling and simulating such flows, as well as present some recent work in the development of a general asymptotic framework that aims to deduce lower-dimensional models that describe the evolution of droplets on surfaces, by bridging the macro- and micro-scale features of the flow. In addition, some preliminary results towards the development of data-driven surrogate models are also presented. Both approaches are shown to adequately capture the dynamics within their domain of applicability, offering new directions and opportunities towards improving our fundamental understanding of this class of phenomena, as well as for informing strategies for controlling and manipulating droplet behaviour in the aforementioned applications and beyond.


Local Competition and Stochasticity as a Nexus for Next-Generation Deep Learning

by Konstantinos P. Panousis, Cyprus University of Technology

Date: Tuesday, 8 February 2022


Deep networks have dominated breakthrough advances in machine learning in the last decade. They have enabled almost human-level accuracy in hard machine learning tasks that were previously considered intractable. The paradigm of deep networks is fundamentally different from previous neural network approaches. Deep networks are statistical machines; they exploit a huge variety, and a great fraction, of the immensely important advances of Statistical Machine Learning of the preceding decade.

Among these, generative modeling and variational Bayes arguments have been extremely impactful. Despite these advances, deep networks face some major challenges: (i) they impose huge memory requirements, with compression techniques being usually relevant only in the limited context of specific network architectures; (ii) they are especially brittle to adversarial attacks; that is, carefully crafted data perturbations, easily recognizable from humans, which, nevertheless, foul the networks into incorrect decisions by exploiting vulnerabilities of the rationale underlying their objective functions; and (iii) they need immense amounts of data to learn a new task, contrary to biological systems which can generalize well given only a single, or few, examples.

In this talk, we present our work on Deep Network arguments that are founded upon the concept of Stochastic Local Competition, incarnated in the form of stochastic local winner-takes-all (LWTA) units. This type of network units results in sparse representations from each network layer, as the units are organized in blocks where only one unit generates a non-zero output. Their main operating principle lies on stochastic arguments, as the network performs posterior sampling over competing units to select the winner. Often, we can combine these LWTA arguments with tools from the field of Bayesian nonparametrics, specifically the stick-breaking construction of the Indian Buffet Process, to allow for inferring the sub-part of each layer that is essential for modeling the data at hand. Such a construction also allows for postulating stochastic network parameters (synapse weights); apart from facilitating generalization, by mitigating overfitting, this approach bears the additional benefit of enabling memory footprint reduction in a Bayesian compression fashion.

Then, training is performed by means of stochastic variational Bayes, with appropriate measures for obtaining low-variance estimators. We show that these arguments give rise to: (i) dense-layer and convolutional networks that completely outperform the state-of-the-art in terms of adversarial robustness to hard benchmarks, without requiring post-hoc processes, such as adversarial training or other data manipulation techniques; (ii) (multimodal) video-to-text networks with self-attention which combine state-of-the-art accuracy with a memory footprint reduction by more than 70%; and (iii) networks that define the state-of-the-art on hard few-shot image classification benchmarks with no compromise in terms of computational efficiency.

In conclusion, we discuss initial results on how these advances can facilitate the goal of obtaining more interpretable deep networks, that transcend the black-box nature of the current paradigm. Progress to this end may catalyze wide technology adoption in products and services, as the latter requires that human experts can obtain an understanding of why, in some cases, a deep learning model generates false positives or negatives.


Computationally Predicting Better Dopants for Higher Mobility Transparent Conducting Oxides

by Prof. David Scanlon, Department of Chemistry, University College London

Date: Tuesday, 1 February 2022


The combination of electrical conductivity and optical transparency in a single material gives transparent conducting oxides (TCOs) an important role in modern optoelectronic applications such as in solar cells, flat panel displays, and smart coatings. The most commercially successful TCO so far is tin-doped indium oxide (Indium Tin Oxide – ITO), which has become the industrial standard TCO for many optoelectronics applications; the ITO market share was 93% in 2013.

Its widespread use stems from the fact that lower resistivities have been achieved in ITO than in any other TCO; resistivities in ITO have reached as low as 7.2 × 10-5Ω cm, while retaining >90% visible transparency. In recent years, the demand for ITO has increased considerably, mainly due to the continuing replacement of cathode ray tube technology with flat screen displays. However, indium is quite a rare metal, having an abundance in the Earth’s crust of only 160 ppb by weight, compared with abundances for Zn and Sn of 79000 ppb and 2200 ppb, respectively, and is often found in unstable geopolitical areas.

The overwhelming demand for ITO has led to large fluctuations in the cost of indium over the past decade. There has thus been a drive in recent years to develop reduced-indium and indium-free materials which can replace ITO as the dominant industrial TCO. In this talk, Prof. Scanlon will outline the strategies that we use in the Materials Theory Group to look beyond the current TCO materials, highlighting the interplay of theory and experiment.


Utilizing Machine Learning for Scale Bridging: From the Atomistic to the Coarse Grained Level and Back 

by Prof. Christine Peter, University of Konstanz

Date: Tuesday, 25 January 2022


Multiscale simulations which combine atomistic and coarse-grained (CG) simulation models can overcome size and time scale limitations of purely atomistic approaches while retaining chemical/biological specificity. In this context, linking the simulation scales and assessing and improving the inevitable shortcomings of the lower resolution models remains an ongoing effort in which machine learning (ML) plays an increasingly important role.

Generally, in bottom-up coarse-graining, CG interactions are devised such that an accurate representation of a higher-resolution (e.g. atomistic) sampling of configurational phase space is achieved. Recently, traditional bottom-up methods have been complemented by machine learning (ML) approaches. ML methods can be used to derive or validate CG models by matching the sampling of a (relatively complex) free-energy surface as opposed to low-dimensional target functions/properties. For example, high-dimensional free energy surfaces can be extracted from atomistic simulations with the help of artificial neural networks (NN) - which can then be employed for simulations on a CG level of resolution. Secondly, ML methods can also be used to obtain low-dimensional representations of the sampling of phase space or to identify suitable collective variables that describe the states and the dynamics of a system.

This information can then be directly fed into the CG potentials or be used to identify optimal CG representations and learn CG interactions. Moreover, the so-obtained low dimensional representations enable us to assess the consistency of the sampling in models at different levels of resolution, to go back and forth between the scales and ultimately to enhance and improve the sampling of the systems.



Dynamics of Bubble Nucleation

by Prof. Carlo Massimo Casciola, Sapienza University of Rome, Italy

Date: Tuesday, 18 January 2022


When the pressure falls below a critical level (cavitation) or the temperature raises above a threshold (boiling), the liquid-vapor transition takes place. The process starts with the nucleation phase, a rare event which is deeply routed in the atomistic nature of the fluid. Successively, depending on the local thermodynamic conditions, the bubble may grow to macroscopic size and couple to the inertial dynamics of the surrounding fluid. In the classical approach each phase is treated separately.

Classical Nucleation Theory (CNT) deals with the nucleation rate (number of bubbles formed per unit time and volume). Once the bubble is formed, the celebrated Rayleigh-Plesset equation, or extensions therein, is classically used to describe the bubble dynamics. After reviewing the state of the art in the field, purpose of the talk is discussing a comprehensive model able to provide a unified description of the different phenomenologies described above. The model is based on the capillary Navier-Stokes equation where the liquid-vapor interface is treated by a diffuse interface model accounting for the relevant thermodynamic properties of the fluid (e.g., equation of state, phase change and latent heat). In order to describe the nucleation phase, a noise term is included (fluctuating hydrodynamics) leading to a system of stochastic partial differential equations with the unique capability of describing the nucleation of vapor cavities from the liquid in the context of continuum mechanics.
Several examples of numerical solutions will be used to illustrate how the model can be effectively implemented on the HPC facilities made available under the auspices of PRACE. They include bubble collapse in free-space and near solid walls and their homogenous and heterogeneous nucleation in different geometries. New results concerning nucleation and bubble dynamics in a flowing liquid and in a confined cavity will also be presented to finally touch upon the rare event techniques aimed at accurately extracting the cavitation pressure for actual water in a wide range of temperatures.


Mechanical Properties of Glassy Polymer Nanocomposites via Atomistic and Continuum Models: The Role of Interphases

by Dr. Hilal Reda, The Cyprus Institute

Date: Tuesday, 7 December 2021


The effect of the properties of an interphase property on the mechanical behavior of the silica–polybutadiene polymer Nanocomposite (PNC) is investigated via combined homogenization and molecular dynamics (MD) simulations. In the current work, we propose a multi-scale computational methodology consisting of detailed microscopic simulations and continuum modelling (homogenization) approaches for predicting the spatial distribution of the mechanical properties in PNCs. The homogenization methodology is based on a systematic nano/micro/macro coupling between detailed atomistic MD simulations and the variational approach based on the Hill-Mandel lemma. The model system glassy-PB/silica NC's was studied at different volume fraction of NP. Using MD simulations, the polymer/NP interphase in PNCs is directly examined by probing the distribution of the density and the stress profile at equilibration.

Based on this analysis, two different interfaces with width of 5 Å and 10 Å, defined from the outer surface of NP, are considered. The effective Young modulus and poison ratio of the interphases are calculated from the relation between the local stress and strain showing high rigidity compared to the bulk material. The mechanical properties at the interphases and the polymer matrix are used, together with homogenization approach, to develop a continuum model for the prediction of mechanical properties of the PNC. A good agreement between the calculated effective mechanical properties through MD and continuum models is observed.



Dynamic Network Models of Seizure Generation: From Theory to Translation

by Prof. John Terry, Interdisciplinary Professional Fellow, University of Birmingham

Date: Tuesday, 2 November 2021


In this talk Prof. John Terry will take us on a journey from scribbled networks on a blackboard in 2012 to the Nature SpinOff final in 2021. He will first introduce mathematical models that were developed to understand the fundamental network mechanisms of seizures. From there he will describe approaches to uncover evidence of network alterations in the background EEG activity of people with epilepsy. Building on this understanding, he will introduce a novel biomarker of seizure risk that is called BioEP and talk through the business development activities were undertaken that culminated in establishing a spin-out company Neuronostics in 2018. He will highlight both successes and failures along the way and reflect on lessons learned.



Post-Processing and Visualizing Large Climate Model Data: Radionuclide Dispersion Modelling and High Performance Computing (HPC)

by Mr Marco Miani, Computational Support Specialist, CARE-C, The Cyprus Institute

Date: Tuesday, 26 October 2021


Visualization is a crucial aspect for analysing and presenting large (giga- or tera-byte) model-generated climate datasets, as the intrinsic complexity of the data structure requires effectiveness and clarity, so to convey the physical significance encapsulated in these data. In this context, special standards (formats) and conventions are used in climate science whose structure and features will presented and addressed in this seminar talk. The exploitation of these formats is a formidable example of efficiency, clarity and interoperability, as well as being a winning strategy to tackle such large-scale problems.

With growing scale, not only geophysical complexity but also computational effort and model-input data procurement grow accordingly: all mathematical models used to investigate climate change and their effects at a global scale, are run and rely on high-performance supercomputers (HPC). Specifically for this talk, an HPC-based atmospheric particle dispersion model, FLEXPART, has been set up and used to simulate a radioactive release from a newly built nuclear power plant on the southern coast of Turkey (Akkuyu), and its consequent dispersion in the EMME region. The technical aspects, as well as the scientific methodology on which these simulations are based and carried out will be presented.



Enalos Chem / Nano Informatics Tools & Isalos Analytics Platform for Drug Discovery and Materials Design

by Dr Andreas Afantitis, Novamechanics Ltd.

Date: Tuesday, 21 September 2021


The development of novel drugs and materials is a multiparameter optimization problem that requires several iterations of designing, synthesis, testing, and analysis. Leveraging artificial intelligence methods and integrating them into automatic chemistry platforms can accelerate the drug/material design cycle. In order to achieve this, we have developed new ideas, methods and algorithms and a range of innovative products and services (e.g., Enalos Suite, Enalos Cloud Platform, chem/nano Pharos database, and Isalos Analytics). The success of drug discovery as well as chemoinformatics-aided material design significantly depends on the success of in silico methods and tools to process, integrate, analyze, and interpret chemical and biological data and properties. The need for efficient data mining and analysis has become very intense especially after the increasing volume of data produced from High Throughput Screening (HTS) experiments. This is a demanding procedure for which various tools must be combined with different input and output formats.

To automate this procedure, it is necessary to bridge automated chemistry, computer-assisted drug/material design, analytics, machine learning (ML) and artificial intelligence (AI). We have developed the necessary tools to facilitate a variety of important tasks to construct workflows that will simplify the handling, processing and modeling of chem/nano informatics data and will provide time and cost efficient solutions, reproducible and easier to maintain. To address this emerging need, NovaMechanics Ltd has developed and integrated within Enalos and Isalos informatics platforms a wide range of Chem/Nano tools, functional within Cloud, KNIME and standalone platform, dedicated to the informatics analysis of chemical and nano data and their corresponding activity/properties. The application of these state-of-the-art chem/nano informatics tools for the development of nanoinformatics models and an in silico drug discovery pipeline for the identification, virtual screening, lead identification and optimization of novel inhibitors as well as the novelty, patent and commercial availability search will be presented.



Polymer / Graphene-based Nanocomposites: A Microscopic View Through the Magnifying Glass of Molecular Dynamics Simulations

by Prof. K. Karatasos, Chemical Engineering Department, The Aristotle University of Thessaloniki, Greece

Date: Tuesday, 7 September 2021


Graphene-based polymer composite systems have recently emerged as novel materials for modern applications in nanotechnology. Such applications include the fabrication of membranes for energy production (e.g, as polyelectrolyte membranes in fuel cells) and for environmental processes (e.g, nanofiltration, desalination etc). Enhancementof the properties of these nanocomposites (i.e, thermal mechanical and electrical behavior) are based on informationrelated tolocal interactions and dynamic processes,that operate down to the atomic level. Parameters such as the presence or absence of solvent, the hydration level, changes in temperature, in pH and in the composition of the membranes, as well as the nature of the interactions at the polymer/filler interface, are among those that play a crucial part in the manifestation of the macroscopic properties of these materials. 

To this end, fully atomistic molecular dynamics simulations can offer a better understanding of the elementary mechanisms associated with the formation of the local microenvironment and the role of different structural features and thermodynamic conditions, contributing thus towards a molecular-level design of polymer/graphene-based composites, with optimized properties.  



Nanoscale Memristors for Neuromorphic Computing Applications

by Dr. Alexandros Emboras, ETH Zürich

Date: Tuesday, 20 July 2021


Technological evolution offered by Moore's law does not progress anymore as it did before.  More and more the high-energy consumption inhibits further progress. New approaches that are more energy efficient and have the potential to solve complex tasks are therefore in demand. The human brain, though, is an example showing that nature can offer much more computational power with much less energy consumption. In this talk I will give an overview of the field of brain-inspired computing and then review our most recent research in the field of opto-electronic memristors for neuromorphic computing applications.


Simulation of Confinement in Nanoporous Alumina

 by Dr. George Papamokos, University of Ioannina




Simulations for Gold Nanoparticles: Electronic Structure, Multi-scale and Data-driven Approaches

by Prof. Ioannis Remediakis, University of Crete

Date: Tuesday, 29 June 2021


Catalysts, sensors, fuel cells, hydrogen storage, batteries, and solar energy conversion are some of the fields where the use of gold nanoparticle shows great promise and presents several challenges. The shape of nanoparticles plays a key role in these applications as it determines the energies of quantum confinement and of surface plasmons, as well as the number and type of active sites for adsorption of ligands and catalysis. The equilibrium shape of metal particles is the polyhedron that minimizes the total surface energy. Such shapes are routinely modelled using variations of the Wulff construction in a multi-scale scheme that combines quantum mechanics and thermodynamics [1].

In this talk, we discuss electronic properties of these nanoparticles in comparison to continuum models [2]. Finally, we discuss the incorporation of gold nanoparticles into polymer matrices, where the properties of the polymer are drastically modified near the polymer-metal interfaces [3].

This work was supported by HFRI project "MULTIGOLD" (HFRI-FM17-1303, KA 10480).

[1] G. D. Barmparis, Z. Lodziana, N. Lopez and I. N. Remediakis, Beilstein J. Nanotechn. 6, 361, (2015)
[2] G. D. Barmparis, G. Kopidakis, and I. N. Remediakis, Materials, 9, 300 (2016).
[3] A. J. Power, I. N. Remediakis, and V. Harmandaris, Polymers 13, 541 (2021).





The Use of Atomistic Simulations to Guide the Derivation and Verification of Molecular Theories

by Assist. Prof. Pavlos S. Stephanou, Cyprus University of Technology

Date: Tuesday, 15 June 2021


Polymeric chains are characterized by a broad spectrum of length and time scales, which give rise to properties that are totally different from those of simple Newtonian liquids. Our aim in this work is to contribute to the understanding of the complex interplay between microscopic chain configurations or conformations and macroscopic behaviour, which is a central goal in polymer science and technology and a prerequisite for the design of improved polymers tailored for specific applications.

Firstly, guided from the reptation theory of de Gennes and Doi/Edwards, we will first show how one can predict the linear viscoelastic properties of polymer melts comprised of long polymer chains by using atomistic trajectories from detailed molecular dynamics (MD) simulations to calculate the primitive path (PP) segment survival probability function ψ(s,t) for entangled melts. This function is a cornerstone of the Doi/Edwards reptation theory but also of all tube models. Direct comparison of the theoretical predictions with the simulation data and previously reported experimental measurements in the literature are in complete accord.

Secondly, we combine MD simulations and the Rouse theory suitably adapted for polymer chains adsorbed by one or both of their ends to offer a quantitative description of local structure and microscopic dynamics in attractive polymer nanocomposite melts using as a model system poly(ethylene glycol) (PEG)-silica nanocomposites. Our work reveals that adsorbed polymer segments in the form of tails and loops on silica exhibit appreciable mobility locally. The simulations also reveal that PEG chains terminated with hydroxyl groups are primarily adsorbed on the silica surface by their ends, giving rise to a brush-like structure, whereas PEG chains terminated with methoxy groups are adsorbed equally probably along their entire contour. Direct comparison of simulation and theoretical predictions, in which information from the MD simulations are used as input, with previously reported experimental data in the literature for the dynamic structure factor for the same systems under the same temperature and pressure conditions reveals excellent agreement.



From Discrete Multiphysics to Deep Multiphysics: How to Combine Particle Methods like DEM and SPH with Artificial Intelligence Using the Particle-neuron Duality

by Dr Alessio Alexiadis, University of Birmingham

Date: Tuesday, 18 May 2021


The algorithm behind particle methods is extremely versatile and used in a variety of applications that range from molecular dynamics to astrophysics. For continuum mechanics applications, the concept of ‘particle’ can be generalized to include discrete portions of solid and liquid matter. This talk shows that it is possible to further extend the concept of ‘particle’ to include artificial neurons used in Artificial Intelligence. This produces a new class of computational methods based on ‘particle-neuron duals’ that combines the ability of computational particles to model physical systems and the ability of artificial neurons to learn from data. During the talk, several applications of these particle-neurons duals will be presented and discussed.




Rediscovering the Most Abundant Protein of Our Body with HPC

by Prof. Frauke Gräter, Head and Scientific Director, “Molecular Biomechanics” Research Group, The Heidelberg Institute of Theoretical Studies (HITS); Professor, Heidelberg University

Date: Tuesday, 04 May 2021


Materials - be it a shoe sole or an Achilles tendon - respond to mechanical stress involving length scales all the way down to atoms and electrons, rendering computational materials science a prime application area of HPC. Prof. Gräter will show how large scale simulations of collagen have fundamentally changed our understanding of this protein, the most abundant building block of our body and its major load-bearing material. We find tension in this biomaterial, as it occurs in our Achilles tendon or muscle, to result in molecular rupture events with vast consequences to the tissue. She will share her recent efforts towards an efficient, scalable and transferable reactive Molecular Dynamics scheme by a combination with kinetic Monte Carlo simulations to cover these and related processes at the relevant time and length scales and yet at atomistic details. The findings, making ample use of HPC infrastructure and in combination with our experimental studies, open new routes towards addressing mechanical stress-induced diseases such as pain or inflammation.


CYGNUS Models for Spectral Energy Distributions of Galaxies

by Prof. Andreas Efstathiou, European University Cyprus

Date: Tuesday, 20 April 2021


Astrophysics concerns the study of the sequence of events that led to the formation of galaxies and the supermassive black holes that usually reside at their centres. Understanding of the complex astrophysics that led to the observed distribution of galaxies, both in terms of properties and numbers, is usually sought within the framework of the standard ΛCDM (Lambda Cold Dark Matter) cosmological model.

Radiative transfer models for galaxies and active galactic nuclei (AGN) are now routinely used for interpreting multi-wavelength observations of galaxies and are increasingly being incorporated into models of galaxy formation and evolution. CYGNUS (CYprus models for Galaxies and their NUclear Spectra) is a collection of radiative transfer models for AGN tori, starburst galaxies and host galaxies.

In this talk, ongoing work funded by the European Space Agency under the PECS program and making use of the CYCLONE facility will be described. The project aims to develop further the CYGNUS collection of models, develop a new method for fitting them to data using a Markov Chain Monte Carlo (MCMC) code and test the method with a large sample of galaxies with excellent photometry and infrared spectrophotometry from the Spitzer Space Telescope. Numerous recent applications of the CYGNUS models as well as prospects for applications with forthcoming missions such as the James Webb Space Telescope (due for launch in late 2021) will be discussed.



Adaptive Algorithms for Schrodinger Type Equations

by Prof. Theodoros Katsaounis, University of Crete, Greece

Date: Tuesday, 13 April 2021


In this talk, Prof. Katsaounis will present a series of results concerning the development and implementation of space time adaptive algorithms for obtaining numerical solutions to linear and nonlinear Schrodinger type of equations. The mathematical foundations of the algorithms are rigorous aposteriori estimates developed for a Crank-Nicolson finite element method for these models. This is a joint work with I.Kyza.



Designing Novel Nanoporous Materials for Applications in Energy and Environment

by Prof. George E. Froudakis, Department of Chemistry, University of Crete, Greece

Date: Tuesday, 06 April 2021


Machine learning techniques (ML) are powerful tools already used in science and industry since their computational cost is by several orders of magnitude lower than that of the “conventional” approaches. However, their ability to provide accurate predictions strongly depends on the correct identification of those parameters (descriptors) that will allow the algorithm to effectively learn from past data. Other critical factors that affect the quality of the predictions are the size and the quality of the dataset used for the training of the algorithm as well as the correct estimation of the training size.
Aiming at both, the transferability of our model and the reduction of the training data set, we introduce 2 different classes of descriptors, based on fundamental chemical and physical properties: Atom Types and Atom Probes. The main difference from previous models is that our descriptors are based on the chemical character of the atoms which consist of the skeleton of the materials and not their general structural characteristics. With this bottom up approach we go one step down in the size of the descriptors employing chemical intuition.

In parallel, an automatic procedure of identifying the appropriate size of the training set for a given accuracy was developed. A novel training algorithm based on “Self-Consistency” (SC) replaced the standard procedure of linearly increasing of the training set. Our SC-ML methodology was tested in 5.000 experimentally made MOFs for investigating the storage of various gases (H2, CH4, CO2, H2S, H2O). For all gases examined, the SC-ML methodology leads to significantly more accurate predictions, while the number of MOFs needed for the training of the ML algorithm in order to achieve a specified accuracy can be reduced by an order of magnitude. In addition, the universality and transferability of our ML model was proved by predicting the gas adsorption properties of a different family of materials (COFs) after training of the ML algorithm in MOFs.

We propose a new methodology for the construction of artificial data (artificial MOFs) with the desired properties that will be used for ML training. This will enable ML algorithms for achieving improved predictions, in particular for high-performing materials. We demonstrate that, after using the artificial data, the capability of the ML algorithms to classify new top-performing MOFs as such, improves remarkably. We are also confident that the present methodology represents an important contribution toward the development of predictive models aiming to the discovery of new materials with outstanding properties. In addition, the main idea of this approach, can be used in many other applications of ML methodologies for overcoming the inherent problem of extrapolation.


Structural Determinants of Function in Complex Microvascular Tissues

by Dr. Igor Chernyavsky, Department of Mathematics; Maternal & Fetal Health Research Centre, University of Manchester, UK

Date: Tuesday, 23 March 2021


Multi-scale structural assessment of biological tissues and organs is essential to gain insight into their structure-function relationship [1,2]. For example, the human placenta is a highly complex and unique multi-functional organ. It is a life-support system that not only nourishes a growing fetus, but also determines her or his life-long health. The placental primary exchange units, terminal villi, host dense networks of fetal capillaries and are interfaced with maternal blood, percolating a disordered porous medium. While placental transport at the micro-scale can be described by established models, systematically upscaling the transport and quantifying the associated uncertainty at the organ-level remain open challenges [1]. This talk will summarise recent progress in advanced 3D microscopy and its assimilation into mathematical models that predict placental function [2]. The models reveal a surprising role of microstructure in the upscaled predictions and demonstrate certain universality of reduced-order approximations for a wide class of solutes and transport regimes [3]. The developed approaches could also be useful for multi-scale modelling of solute exchange in other complex microvascular systems.


Synergy of HPC and Nuclear Physics to Resolve Long-standing Puzzles

by Assist. Prof. Martha Constantinou, University of Temple in Philadelphia, USA

Date: Tuesday, 16 March 2021


More than 99% of the mass of the visible matter resides in hadrons which are bound states of quarks and gluons, collectively called partons. These are the fundamental constituents of Quantum Chromodynamics (QCD), the theory of strong interactions. While QCD is a very elegant theory, it is highly non-linear and cannot be solved analytically, posing severe limitations on our knowledge of the structure of the hadrons. Lattice QCD is a powerful first-principle formulation that enables the study of hadrons numerically, which is done by defining the continuous equations on a discrete Euclidean four-dimensional lattice.

Hadron structure is among the frontiers of Nuclear and Particle Physics, with the 2015 Nuclear Science Advisory Committee’s Long Range Plan for Nuclear Physics identifying a future electron-ion collider (EIC) as the highest priority for new facility construction. Last year, the National Academies of Sciences, Engineering, and Medicine (NAS) released an assessment report which strongly endorses the science case for an EIC. The NAS report identified three high-priority science questions to understand hadron structure:

1. How does the mass of the nucleon arise?
2. How does the spin of the nucleon arise?
3. What are the emergent properties of dense systems of gluons?



Active (Non-) Particles: Donuts, Curved Rods, and Flexibility

by Dr. Thomas Montenegro-Johnson, Senior Lecturer in Mathematical Biology, University of Birmingham

Date: Tuesday, 9 March 2021


Fun things start to happen when active particles get kinky. This talk will focus on what happens when active particles move from point-like geometries such as spheres, into non-trivial topologies such as tori, and slender 3D curved structures, offering opportunities for rich dynamical behaviours and new control strategies exploiting shape changes and flexibility.



Special Displacement Method for the Calculation of Materials’ Properties at Finite Temperatures

by Dr. Marios Zacharias, Post-doctoral Researcher, Department of Mechanical and Materials Science Engineering, Cyprus University of Technology

Date: Tuesday, 16 February 2021


Typical first-principles (parameter free) calculations of the optoelectronic properties of solids are performed by describing the nuclei as classical particles clamped at their crystallographic positions. This approximation undermines the accuracy of predicting materials properties at, or above, room temperature, since it misses quantum nuclear and thermal effects [1]. Recently, an ex-novo approach, namely the special displacement method (SDM) [2,3], has been developed to include these effects in state-of-the-art electronic-structure calculations.

In this talk, the speaker will demonstrate how high-performance computing in conjunction with SDM can advance the field of electronic-structure calculations to explore novel emergent phenomena arising from the interplay of electron-phonon interactions. He will present some applications of SDM, including the calculation of full temperature-dependent band structures and optical spectra of crystalline and nano-crystalline quantum systems [2,3,4]. SDM holds promise for high-throughput calculations of many material’s properties at finite temperature.



Multiscale Modeling of Biomolecules and Materials

by Prof. Zoe Cournia, Biomedical Research Foundation, Academy of Athens

Date: Tuesday, 02 February 2021


In this webinar, the method development and applications of multiscale computational techniques for the modeling of materials and biomolecules ranging from atomistic and coarse-grained Molecular Dynamics (MD) simulations, Monte Carlo, Markov state models, and Machine Learning with applications in drug design and drug delivery design will be discussed. Biological membranes comprise fascinating examples of soft matter interfaces involved in a wide range of biological and industrial functions. Several aspects of the structure and dynamics of biomembranes as well as methodologies for targeting specific membrane interfaces for developing novel drug candidates and nanoparticle delivery systems will be presented.



Computer Simulation Aspects of Nanoparticle and Nanodevice Design 

by Prof. Panagiotis Grammatikopoulos, Staff Scientist, "Nanoparticles by Design” Unit leader at OIST Graduate University, Japan; Visiting Assistant Professor at the Particle Technology Laboratory at ETH Zürich, 

Date: Tuesday, 26 January 2021


Cluster beam deposition (CBD) is a term that collectively describes various physical methods of nanoparticle synthesis by nucleation and growth from a supersaturated atomic vapour. It provides a solvent- and effluent-free method to design monodisperse multifunctional nanoparticles with tailored characteristics that can be subsequently deposited on a desired substrate or device in the soft-landing regime under ultra-high vacuum.

In this talk, I will explain the main mechanisms that control the basic properties of individual nanoparticles such as size, shape, or chemical ordering, based on various setups of CBD sources. Moving to a coarser scale, I will bring up examples where larger structures can be designed using nanoparticles as their functional building blocks, such as novel sensors and energy storage devices. To date, CBD faces two main limitations that need to be overcome for real-world applications: (i) limited yield, and (ii) precise structural control. The main thesis of this talk is that both challenges can be tackled by in-depth theoretical understanding of both the thermodynamics and kinetics of nucleation & growth. To this end, atomistic computer modelling can be an invaluable tool, complementing experimental fabrication and guiding future source design.



High Performance Computing in Structural Biology
by Assist. Prof. Vangelis Daskalakis, leader of the Computational Environmental Modeling (CEM) Group, Department of Environmental Science and Technology (EST) at the Cyprus University of Technology (CUT)

Date: Tuesday, 19 January 2021


Structural biology is a branch of molecular biology, biochemistry, and biophysics with the focus on the molecular structure of biological macromolecules (like proteins and membranes). A key question of Structural Biology is how these macromolecules acquire their structures. A combination of methodologies like Molecular Dynamics and Markov state modeling that is supported by High Performance Computing (HPC) leads to an ideal description of macromolecule conformations and dynamics at all-atom resolution. The latter is a valuable information in Structural Biology. In this seminar two key HPC projects will be presented that relate to: (A) the out-break of the novel coronavirus (SARS-CoV-2) that causes the respiratory tract disease COVID-19 and the dynaics of two key viral proteins that can potentially be used as drug-targets or source of epitope vaccines, and (B) the delicate balance between light harvesting and photoprotective modes of the Light Harvesting Complexes of Photosystem II, that can potentially be used as models for artificial photosynthesis.



Fostering innovation with HPC, advanced simulations and AI and Big Data
by Dr. Christos Christodoulou, Managing Coordinator - Innovation Scout, SimEA, CaSToRC, The Cyprus Institute

Date: Tuesday, 12 January 2021


The computation-based science and technology research center (CaSToRC) has recently become a National Competence Center in HPC and the host of the prestigious SimEA ERA Chair project that aims to expand the research activities of the center and forge innovative collaborations with industry.

In this talk, Christos, the Innovation Scout of SimEA, will talk about the Management and Innovation Office (MIO) established at CaSToRC, as a structural change aiming to promote the innovation strategy of the center. He will also talk about the core competences of the SimEA team and of CaSToRC, and how the industrial and governmental sector can use HPC to achieve breakthroughs in areas like energy and environment, design of novel materials, digitalization of industry, circular economy, oil and gas production, medicine, and financial risk assessment.




Presentations by our PhD students who have recently joined the SimEA research team. Here, they present their MSc research projects and give an overview of the research that they will be conducting during their PhD at The Cyprus Institute.

Date: Tuesday, 8 December 2020

  • The Optimization Algorithm in Machine Learning
    by Eleftherios Christofi, PhD Student, SimEA project, The Cyprus Institute In this weminar


The last two decades have marked a rapid and significant growth of the Artificial Intelligence field. Deep learning using artificial neural networks became an essential tool for a vast number of applications fields. The structure of deep learning relies on basic concepts from several mathematical fields, such as linear algebra, calculus, optimization, statistics. This presentation is an introduction to the mathematical background of deep learning. In particular, we focus on the optimization algorithm widely used, namely the stochastic gradient descent. We study and compare the behaviour of different variants of this algorithm under various circumstances and summarize their strengths and weaknesses. The goal of this presentation is to provide the viewer with the knowledge to comprehend the training procedure of a neural network.


  • Development of an algorithm and software for computing Pareto fronts with a gradient-based method, with applications in aerodynamics
    by Nikolaos Patsalides, PhD Student, SimEA project, The Cyprus Institute


This diploma thesis proposes, develops and evaluates a low-cost algorithm and software for computing Pareto fronts with a gradient-based method, i.e. a method that uses the gradient of the cost functions, by implementing the algorithm in aerodynamic shape optimization. It is based on a method that successively computes points on the front, by properly forcing all objectives to equality constraints, apart from one that becomes the target to be minimized. After the description of the algorithm's steps, applications in aerodynamic shape optimization of an airfoil are presented. Two- and three-objective problems are solved, with bounded design variables and equality or inequality constraints.



Webinar: Bottom-up Approach - From Model Molecular Systems to Complex Polymer Materials
by Dr. Petra Bacova, Post-doctoral researcher, SimEA project, The Cyprus Institute

Date: Tuesday, 24 November 2020


Following the lecture of Prof. Harmandaris, a postdoctoral SimEA fellow Dr. Petra Bacova will shed further light on the computational design of advanced materials. The main focus will be on the complex polymer materials, addressing three main questions: WHY these materials are interesting, WHAT their composition is and HOW we can model their properties computationally.



Commencement of seminar series - inaugural lecture by the ERA chair Prof V. Harmandaris

Date: 5 November 2020

Title: Computational Science and Engineering of Complex Materials: The SimEA ERA Chair Initiative


The ERA Chair project “Modeling and Simulation for Engineering Applications” (SimEA), funded by EU, refers to a new initiative at CaSToRC of CyI, regarding the challenges and opportunities arising when advanced computing and data science are utilised to solve engineering problems. The research team, led by ERA Chair Prof. V. Harmandaris, will work on the development of mathematical and computational methodologies for complex molecular systems, with important applications in nano/bio technology.

The team will pursue a program of research excellence and innovation by applying and developing mathematical and computational methods, including multiscale modeling, physics-based and/or data-driven molecular models, uncertainty quantification, and machine learning methods, integrated with High-Performance Computing, for tackling challenging problems in different application areas related to Computational Science and Engineering.  Examples include a broad range of systems/materials of great scientific and technological interest, such as nanocomposites, polymers, graphene-based nanostructured materials, proteins, and biomolecular systems.

In this inaugural seminar of the seminar series under the SimEA, we will give an overview of the main objectives and research topics of the SimEA. Our vision is to create a regional and national hub for advanced computing and its applications at CyI, collaborating with academia and the private sector, across several research areas, such as: (a) Multi-scale Modelling and Simulations of Nanostructured Materials, (b) Data-driven Machine Learning Approaches for Modelling Across Scales, (c) Modelling of Biomolecular Systems for Biotechnology Applications, (d) Statistical Inference for Uncertainty Quantification and Model Selection.