|
Department of Materials Science and Technology, University of Crete www.materials.uoc.gr | Hellenic Society for the Science and Technology of Condensed Matter www.hsstcm.eu | |
Department of Physics, University of Crete, Greece and Institute of Electronic Structure and Laser (IESL), FORTH.
.
Department of Chemistry, University of Crete, Heraklion, Greece
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 of the dataset used for the training of the algorithm (commonly called training set) 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 [1] and Atom Probes [2]. The main difference from previous models is that our descriptors are based on the chemical character of the atoms which consist 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. On parallel, an automatic procedure of identifying the appropriate size of the training set (TS) for a given accuracy was developed [3]. A novel training algorithm based on “Self-Consistency” (SC) replaced the standard procedure of linearly increasing of the TS (100,200,300…). Our SC-ML methodology was tested in 5.000 experimentally made MOFs [4] for investigating the storage of various gases (H2, CH4, CO2, H2S, H2O). For all gases examined, the use of both descriptors instead of building blocks 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. Finally, 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.
Department of Materials Science, School of Natural Sciences, University of Patras (galanakis@upatras.gr)
All-electron ab-initio electronic band structure methods are a powerful tool to study the magnetic properties of materials with various applications in modern nanotechnology. Such computational methods have been extensively used both to predict novel phenomena and materials as well as to understand the physics behind experimental discoveries and observations. We will discuss the fundamentals of these methods and the properties of interest to compute (like spin moments, magnetic anisotropy energy, magnetic spectral properties, magnetic order, Weyl nodes etc) and how the computed results compare to available experimental data. These results can form the basis for new theories and models focusing on temperature dependent properties of the magnetic materials like Curie temperature, spin-waves (magnons) and the behavior of the magnetization as a function of the temperature. In all cases examples of realistic calculations based on ell-electron ab-initio methods will be provided.
Department of Materials Science, School of Natural Sciences, University of Patras
Using a combination of first principles and semi-empirical calculation we explore the structural, electronic and optical properties of a wide range of perovskite (ABX3) nanoparticles of different sizes, shapes and composition. The variations of the BX3 backbone structure considered, include all possible combinations of the cations B=Pb,Sn and the anions X=Cl, Br, I, while the interstitial cation A is either methylamonium (MA), or formamidinium (FA), or guanidine amine (GA), or dimethylamine (DEA). In all cases the nanoparticles' geometries were thoroughly relaxed, and their stability was examined by comparing their binding energy per unit formula. Their optical properties were calculated by means of RT-TDDFT calculations which gives a more complete picture of the absorption spectrum (not just a few low lying excited states, as in linear response calculations). The results clearly demonstrate the effect of quantum confinement on the properties of the nanoparticles.
Department of Mathematics and Applied Mathematics, University of Crete and Institute of Applied and Computational Mathematics, FORTH
Macromolecular systems (polymers) are characterized by a very broad range of characteristic time and length scales; for example, even a single polymer chain exhibits length scales ranging from the bond length (~ Å) to the size of the chain (O(10 nm)) and corresponding time scales ranging from a few fs (10-15 sec) for the bond vibrations up to the order of milliseconds or even seconds for the whole chain relaxation. For this reason, the computational study, and the prediction of structure-property relations, of polymer-based materials is a very challenging, and intense, research field. Due to the above spectrum of spatiotemporal scales, a systematic combination of simulation approaches, describing a specific system in different levels, is necessary [1,2]. Here, we give an overview of hierarchical computational approaches, involving all-atom and mesoscopic (coarse-grained, CG) dynamic simulations for polymer-based materials. Atomistic molecular dynamics simulations offer a direct approach for the detailed study of systems, in the classical approximation. However, due to the computational costs it is not feasible to take into account all the atomistic degrees of freedom in long, high molecular weight, polymer chains. To reduce the dimensionality of the model systems, and increase the length and time scales accessible by simulations, coarse-grained models that describe group of atoms as “superatoms”, or “particles”, can be used. We describe systematic “bottom-up” CG models, which are obtained by applying data analytics approaches on large datasets (configurations) obtained from atomistic simulations, providing also error estimates for the derived CG models [2,3]. In addition, we present machine learning based approaches for linking different scales, by re-inserting atomic detail in CG configurations [4]. The systematic combination of the above approaches allows the study of macromolecular systems, of high molecular weight, over a broad range of time scales, from a few fs up to ms. and the prediction of their (structural, dynamical, rheological, etc.) properties. As examples, we present results concerning the properties of various polymer-based materials; from polymer melts, up to polymer thin films and graphene-based polymer nanocomposites [5-8].
Department of Materials Science and Technology,University of Crete and Institute of Electronic Structure and Laser (IESL), Foundation for Research and Technology Hellas (FORTH))
Metamaterials are artificially structured materials with novel and unique electromagnetic properties arising mainly from the shape and distribution of their subwavelength-scale building blocks. Arranging properly those building blocks one can achieve properties such as negative permeability (even in the optical region), negative refractive index, extreme permittivity and permeability values, giant chirality, unusual anisotropy etc. All these properties provide a unique vehicle for the control of electromagnetic waves, and can be exploited in a variety of applications, including imaging, sensing, telecommunications and information processing, etc. In this talk I will review some of the recent metamaterials-related activities of our group, emphasizing on chiral metamaterials, i.e. metamaterials lacking any symmetry plane. Such metamaterials are able to give giant and engineerable optical activity and circular dichroism, as well as negative index of refraction at a desired frequency band. Furthermore, I will discuss our recent efforts to combine properly chiral metamaterials with gain media as to obtain chiral parity-time (PT) symmetric systems. PT-symmetric systems are associated with novel and unusual properties, such as unidirectional invisibility, simultaneous coherent perfect absorption and lasing, etc. As I will discuss in the talk, combining PT symmetry with chirality promises novel devices, combining all the features of PT-symmetric systems with desired output wave polarization.
Department of Physics, University of Thessaly
Fluids play an important role in several phenomena and applications such as water desalination, water purification and drug delivery, just to mention few ones. The development of microfluidic and nanofluidic devices with applications in the above fields make their study increasingly important. However, as the size of fluid systems goes down to micro and nanoscale, various phenomena and properties are significantly affected. This effect should be taken into account in the design of complex fluid systems incorporating various length scales. In order to put in evidence size effects we employ appropriate simulation methods: Molecular Dynamics Simulations (for nanoscale), Dissipative Particle Dynamics (for mesoscale). The results show that as the size of fluid channels gets small, transport properties and slip are significantly affected by the size of channels, nature of walls (hydrophobic, hydrophilic) as well as wall roughness and corrections to macroscopic laws should be applied in conventional macroscopic laws in order to achieve appropriate design.
Environmental Research Laboratory (EREL), National Center for Scientific Research “Demokritos” (NCSRD)
The exponential growth of computational power offers significant opportunities for more detailed studies of physical systems. The simulation of the physical world could be categorized based on the spatial and time resolution under study, as (a) macroscopic, (b) mesoscopic and (c) microscopic. The latter consist of molecular simulation studies, whereas the physical world is described by the average behavior of molecular interactions. As a result, the phase space is explored by deterministic or stochastic algorithms and thermodynamics properties are calculated by using statistical mechanics tools. In addition, the big computational power permits the production of “Big Data” and as a result, the necessity to process and analyze them, is emerged. Machine Learning algorithms are data analytics tools used in many fields and in the case of Natural Sciences, such as molecular simulations and material design. The process of studying a system is by far extended. No equation or model that describes the system exists, and the goal of the study is to deduce (“learn”) the model from the data.
Department of Physics, Aristotle University of Thessaloniki, GR-54124 Thessaloniki, Greece
Clusters, aggregates of atoms with sizes typically ranging from sub-nanometer to few tens of nanometers, exhibit size, shape, structure and composition dependent properties different from their bulk counterparts. Core/shell nanoparticles(NPs), in which the core and the shell are of different materials, offer a way to relax the strain induced by the lattice parameter mismatch and create defect-free interfaces, owing to their large surface to volume ratio. Moreover, numerous structural phenomena related with metallic NPs such as coalescence, sintering, segregation and aggregation are frequently observed and modeled using Molecular Dynamics (MD) and Monte Carlo (MC) methods. In addition, NP-surface interaction involves a variety of fundamental phenomena, providing a unique system for studying electron transport mechanisms. Ab initio as well as interatomic potential based MD and MC simulations are employed to investigate the structural, thermal and electronic properties of a variety of metallic NPs. The band structures and charge transfer of the nanoparticles supported on metal oxides are scrutinized in several cases through ab initio simulations of thousands of atoms.
Theoretical and Physical Chemistry Institute, National Hellenic Research Foundation, Vass. Constantinou 48, 11635 Athens, Greece
Membranes based on nanoporous graphene consist one of its most promising applications. In this presentation, we present how relevant quantities like the energy barriers (activation energies) and permeance can be calculated with first-principles electronic structure methods. We then present our theoretical results on the permeation of several molecular systems through pores in single layer graphene with the goal to determine the size and type of pores with optimal permeability and selectivity. Our study was performed at the level of DFT (hybrid-meta GGA functionals). We particularly focused on pores that are created by carbon vacancies and nitrogen doping (pyridinic defects). We demonstrate that the size of interest for gas separation is 0.5 nm and show examples of pores with industrially acceptable permeance that can effectively separate gases. Finally, we turn our attention to pore stacking in bilayer graphene which are studied with atomistic simulations. We show that combinations of pores can be used to control molecular permeability.
Computational Materials Science Laboratory Department of Materials Science and Engineering, University of Ioannina, 45110 Ioannina, Greece Institute of Materials Science and Computing, University Research Center of Ioannina, 45110 Ioannina, Greece
Graphene’s optical response is characterized by constant absorption in the visible but electrically-tunable absorption in the NIR-SWIR and plasmonic excitations in the midIR-LWIR spectrum. These traits make for interesting applications in photodetection, light modulation and sensing. To make the response more efficient and competitive, however, the small overall absorption in graphene must be overcome by integrating graphene with resonant photonic or plasmonic cavities. Strong light absorption within the resonators creates hot electrons and temperature gradients. In a comprehensive modeling and design scheme of graphene-based optoelectronic applications, the optical, thermal and electrical responses must be considered within a self-consistent approach: absorption creates hot carriers, whose temperature distribution is determined by the thermal properties of graphene and the appropriate relaxation pathways and corresponding rates. But the thermal properties and the absorption in graphene are themselves functions of the temperature. To successfully solve this, self consistent calculation cycles are employed until convergence. After a short introduction in computational methods in plasmonics and graphene, we will explore several approaches for graphene-based optoelectronic devices.
Institute of Applied and Computational Mathematics, FORTH and Department of Mathematics and Applied Mathematics, University of Crete
Most real world systems include a description at several scales, i.e., they are multi-scale in nature. Our understanding in models from physics is by far the best, however, even there, many issues remain mathematically unexplored or inaccessible by computational means. In several complex applications involving multiple scales, notably in materials science and nanotechnology, the closure laws are either not known, or hold in restrictive situations. From a computational prespective, the huge number of degrees of freedom of the microscopic problems, as well as, the possible singular behaviour of the underlying phenomena constitute the main bottlenecks for important developments. Usually, effective theories fail when singular phenomena appear, e.g., the passage from atomistic to continuum is typically valid only when certain smoothness criteria are met. A way to overcome this problem at a computational level is to use hybrid models across scales, i.e., to use different models in different areas of the computational domain. We would like to study models where the microscopic model is kept in the areas of singular behaviour and macroscopic models are used in the areas of smoothness. Numerical Analysis motivates the design of new consistent models which accurately link the two scales of the problem. In this talk we will see how one can address this problem when coupling between atomistic and continuum descriptions is chosen.
Department of Physics, National and Kapodistrian University of Athens, 15784 Panepistimioupolis, Athens, Greece
The spin orbit torque effect [1] provides an efficient way of manipulating the magnetization of ferromagnets. In response to an electrical current, the magnetization is subjected to a rotation, eventually reaching the opposite direction. In this process, a fundamental role is played by the spin-orbit coupling and the absence of space-inversion symmetry. These two properties are shared by topological insulator surfaces, resulting in host metallic states with spin-momentum locking. In such systems, it has been shown that magnetic defects may interact, resulting in a ferromagnetic state. Thus, they suggest themselves as ideal for the spin-orbit torque effect. Employing density-functional theory, a multiple scattering formalism and the semi-classical Boltzmann transport equation, we present calculations on ferromagnetically coupled magnetic transition-metal defects on the surface of the topological insulator Bi2Te3. We find a strong spin-orbit torque acting on the defect magnetic moments as the combined result of the special features of the Bi2Te3 surface states and the relative orientation of the magnetic moments with respect to the Fermi-surface spin polarization [2].
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (No. 1314). The authors acknowledge computational time granted from the Greek Research & Technology Network (GRNET) in the National HPC facility – ARIS – under project ID pr00504_thin-TopMag.
Department of Materials Science, University of Patras
Phononic crystals (PC) are artificially made materials. Under certain conditions, they have frequency regions (the so called phononic band gaps, PBGs) where the propagation of elastic waves is prohibited. PBGs may appear in different frequencies depending on the dimensions of the PC. So, for PC in the nanometer scales, PBGs appear at the THz region while for PC in the meter scales, their PBGs appear at the Hz region. In this presentation, the PC and their features and application will be shown. At the nanoscale PC can be used to control thermal conductivity and have recently been proposed for phononic communications. At the sale of micrometers, they can be used as sensors. Finally, for PC with dimensions in meters can be used to control vibrations (e.g. seismic waves). Emphasis will be given in the computational methods used at the different length scales. Although they have been studied for almost 30 years, they are still of interest in their extreme dimensions. For dimensions in nanometers, the interest is from Materials Scientists, Physicists and Chemists, while for dimensions in meters, the interest comes from Civil and Mechanical Engineers.
Department of Physics, National Technical University of Athens
Already for half a century, Density Functional Theory (DFT) calculations have been among the most widely used methods in the fascinating and ever evolving field of materials modelling. The versatility and accuracy of the DFT approach (with its myriad of variants), together with the astonishing increase of computational power available for simulations, have made such calculations an indispensable tool in explaining available experimental data for a wide range of problems in Physics, Chemistry and Materials Science, but also in proposing new systems and phenomena suitable for various applications. In this talk, after discussing the fundamentals of the DFT on exploring the properties of materials at the atomic-scale, a number of examples from recent or ongoing research studies will be used to demonstrate how first principles quantum mechanical calculations play its own key role in the development of devices at the forefront of technology fields that include, for example, electronics, optoelectronics and laser printing.
Webpage credits: Ioannis Remediakis and Niki Sorogas