Computer modelling in nanoscience and nanotechnology: an atomic-scale perspective III

Resume : I will describe our development of first-principles (FP) model potentials (MPs) for large-scale lattice-dynamical simulations. Our MPs are applicable to cases in which the atoms in the material preserve their basic connectivity or bonding topology. This allows us to identify a suitable reference configuration and parametrize the energy as a Taylor series for all the atomic distortions with respect to the reference state. Such a textbook approach has many advantages: the MPs are trivial to formulate for any material and physically transparent, the involved approximations are clear, and the accuracy can be improved in a systematic manner. Further, because the energy depends linearly on the MP parameters, we can adopt efficient strategies for their fitting, obtaining a very fast and robust scheme. Also, at variance with the usual perturbative approaches, we express the energy in terms of products of displacement differences, so that translational invariance is guaranteed by construction and the parameter fitting gets further simplified. I will illustrate our scheme with especially challenging cases, namely, ferroic ABO3 perovskite oxides (including nanostructures) undergoing transitions driven by soft phonon modes, which today attract great interest for novel energy-harvesting applications, among others. I will show how our ability to run large-scale simulations with FP accuracy has allowed us to reveal surprising effects that had remained hidden to previous approaches.

Resume : In the computer modelling for either nano- or bio-science, free-energy has become a significant realistically-reachable quantity for describing thermodynamic properties. It is very important to develop a method that enables us to construct free-energy profile without empirical force-fields. Meanwhile, some shortcomings of thermodynamics integration method, which is currently a standard method for constructing free-energy profiles, have been pointed out; poor statistical sampling resulting from a breakdown in the ergodicity, numerical integration as a post-processing, and inconvenient analysis for the multi-dimensional collective variables (reaction coordinates). We have developed a mean-force dynamics approach to construct free-energy profile without empirical potential parameters by combining the logarithmic mean-force dynamics (LogMFD) [1] and first-principles molecular dynamics, calling this new method first-principles LogMFD (FP-LogMFD) [2]. This allows us to sample higher energy states efficiently, to evaluate the free energy at the local point of reaction coordinates without any post-processing, and furthermore to do with keeping a level of accuracy in the first-principles approach. [1] T. Morishita et al., Phys. Rev. E 85, 066702 (2012); J. Comp. Chem. 34, 1375 (2013). [2] M. Nakamura M. Obata, T. Morishita, and T. Oda, J. Chem. Phys. 140, 184110 (2014).

Resume : Electrons in rare-earth magnets can be classified as two types: itinerant electrons and localized ones. Fe 3d as well as Nd 5d electrons are examples of the former, while Nd 4f electrons are the latter. In particular, itinerant magnetic states dominate interactions among magnetic sites directly. Thus, itinerant d states are expected to be sensitive to lattice strain, although the current status of the understanding on this issue is being far from complete, in particular for the magnetic anisotropy. In this work, we have investigated strain effects on magnetic properties on the basis of density functional theory for Y2Fe14B, where Y is a prototypical f0 rare earth element [1]. First-principles calculations are performed using the OpenMX code. We changed the lattice constants from the equilibrium values in various manner. We found that the uniform compression enhances the perpendicular magnetic anisotropy. To clarify the origin of this enhancement, we developed a new method to decompose the magnetic-anisotropy energy into contribution from each atomic site as well as from couplings among specific atomic orbitals. This method employs second-order perturbation theory and an on-site approximation for the spin-orbit coupling with the form of l dot s neglecting any off-site interactions. As a result, we clarified that the Fe j2 site plays a significant role in enhancing the magnetic anisotropy. [1] Z. Torbatian, T. Ozaki, S. Tsuneyuki, and Y. Gohda, Appl. Phys. Lett., in press.

Resume : Graphene provided the excitement and impact to explore isolated one-atom thick layers that go beyond graphene. The boron-carbon-nitrogen 2D layers are now widely explored since exhibit a variety of structural and electronic properties that depend on the composition. In this work, we develop strategies for predicting the most stable 2D BxCyNz layers. The idea is to find 2D materials that preserve the honeycomb-like structure and can complement graphene in properties. The modeling of these structures is done using the cluster-expansion method combined with extensive first-principles calculations. Doping of graphene with carbon and boron atoms was studied before, using both first-principles and cluster-expansion methods. The purpose of those studies was mainly to investigate the induced by doping band-gap opening in graphene. For instance, for the Cx 1Nx system two stable ordered semiconducting structures, C12N and C3N (x = 0.08 and 0.25, respectively), have been predicted through the cluster-expansion technique (C12N has a direct band gap of ~1 eV). In this work, we explore systematically the stability and properties of 2D BxCyNz layers with a full spectrum of compositions. For the mentioned above Cx-1Nx system, we predict, for instance, that the most stable layer is obtained for x = 0.66. Interestingly enough, our calculations reveal that the 2D layer splits into 1D stripes. The nanoribbons are not connected with each other (except for weak interactions) and have a band gap of 2.9 eV. Furthermore, the building block of each nanoribbon is a C2N4 molecule with a 1.1 eV/molecule energy gain upon formation of the stripe.

Resume : Many technologically important, chemically complex phenomena are currently beyond the reach of standard or O(N) first principles molecular dynamics (MD) techniques because the necessary model system sizes are too large, and/or the required simulation times are too long. In most situations, using classical MD is not a viable alternative, as suitably general and accurate “reactive” force fields are not available, nor are fitting databases a priori guaranteed to contain the information necessary to describe all the chemical processes encountered along the dynamics. In this talk I will argue that this situation forces the use of MD techniques capable of incorporating accurate QM information generated at run time during the simulations [1, 2]. It also effectively creates a novel market for dynamical databases coupled with specially-tuned “Machine Learning” force fields which minimise the computational workload by allowing QM subroutine calls only when “chemically novel” configurations are encountered along the system’s trajectory. I will present one such “Learn On the Fly” scheme, effectively unifying First-Principles Molecular Dynamics and Machine Learning into a single, information efficient simulation scheme capable of learning/predicting atomic forces through Bayesian inference [3]. References [1] J.R.Kermode, L.Ben-Bashat, F.Atrash, J.J.Cilliers, D.Sherman and A.De Vita, Nat. Commun. 4, 2441 (2013). [2] A. Gleizer, G. Peralta, J. R. Kermode, A. De Vita and D. Sherman, Phys. Rev. Lett., 112, 115501 (2014). [3] Z. Li, J. R. Kermode and A. De Vita, submitted.

Resume : Here we discuss how to extend the domain of application of the Kinetic Lattice Monte Carlo to the large scale (micron for the space and minute for the time) atomistic simulation of system's kinetics characterized by an inner structural transition from one to another crystal structure. The feasibility of this extension is demonstrated for the interesting process of the epitaxital graphene synthesis on silicon carbide (SiC) substrate by selective evaporation at high T of Si atoms. The main steps governing this synthesis process are the following: 1) defective sites are the preferential regions of the sublimation phenomenon, 2) C atoms migrate after the breaking of the Si-C tetrahedral network of bonds and reconfigure in the graphitic lattice after a densification, 3) the densification is mediated by not-ordered weakly bounded configurations, 4) the process can continue leading to the formation of new grephene layers. The proposed method exploits a scaffold composed by inter-penetrating three dimensional (3D) lattices for the characterization of the system's micro-state. Simulated kinetics proceeds by intra-lattice and inter-lattice activated transition events involving single atoms. In the particular case of study, we demonstrate that quantitative predictions of the process evolution as a function of the initial state and the process's parameters can be obtained and readily compared with experimental analyses of processed samples.

Resume : Nanomaterials attracted much attention from materials researchers in the last few decades due to their unique properties. Graphene is a good example of nanomaterial which is considered one of the most promising for developing nanoelectronics in the near future. There is also interest in an inorganic counterpart of graphene, the hexagonal Boron Nitride (h-BN) which is composed of alternate boron and nitrogen atoms in a hexagonal array sheet. In this context, there is a special interest in a hybrid material, which is composed by h-BN lattice with random substitution of some B-N atoms by carbon atoms, the so called h-BNC. Experimentally it is known that large sheets of h-BNC constituted by h-BN and C, distributed randomly, are promising candidates in the electronics area, due to its chemical stability, electronic and mechanical properties. In the present work, we performed a series of simulations for h-BNC structures using molecular dynamics (MD) techniques with reactive force fields (ReaxFF) and also using the Tight Binding Density Functional Theory (DFTB) in order to estimate several mechanical properties for these systems at various C concentrations. Young modulus and poisson coefficient are shown to vary with C substitution percentage from 0% up to 100%. Differences between ReaxFF and DFTB results are discussed.

Resume : Poly(dimethylsiloxane) (PDMS) is the most popular elastomeric material since it couples biocompatibility with excellent elastic properties. For this reason considerable efforts are currently being concentrated on the fabrication of stretchable metallic circuits and microelectrodes integrated on PDMS. Recently it has been demonstrated that neutral metallic nanoparticles produced in the gas phase and aerodynamically accelerated in a supersonic expansion can be implanted in PDMS substrate to form a conductive nanocomposite (nc) with novel functional properties[C. Ghisleri et al., J. Phys. D: Appl. Phys 47,015301 (2014)]. While the effectiveness of this novel technique has been already demonstrated few information are available on the ncs elastic properties as a function of the implanted clusters concentration. To this aim we perform dynamic mechanical simulations to estimate the ncs Young modulus for nc samples having different cluster concentrations. The results show that the nanocomposite Young modulus is basically unaffected (with respect to pristine PDMS) up to a cluster concentration of 25%; while above 25 % we observe a Young modulus exponential increase. These results agree well with nanomechanical data based on atomic force microscopy and demonstrate that SCBI can be used to produce metal/polymer nanocomposite materials with reproducible mechanical properties.

Resume : Molecular dynamics (MD) is a statistical mechanics computational approach that provides the opportunity to access basics phenomena involved in the heat transfer at the nanoscale. MD is generally used either to extract bulk conductivities from the heat current fluctuations during a NVE simulation (Green-Kubo or EMD approach), or bulk conductivities and interface resistances from the temperature profile once the stationary regime between a hot and a cold reservoir is reached (direct method or NEMD). We have developed an alternative framework of MD simulations for the study of thermal conductivities and interface resistances. The method, called approach-to-equilibrium MD (AEMD), relies on the Fourier-based exploitation of the transient regime to equilibrium, i.e. flat temperature profile. It is computationally faster than the two other approaches and can be applied to a range of systems. We will present in particular application to bulk Si, Ge and a-quartz and discuss the length dependence originating from micrometric phonon mean free paths in Si. We will also propose different solutions to exploit the temperature transient in the case of interfaces and nanoconstrictions depending on the weight of the interface resistance compared to the conductivity of the materials at both side. Finally we will open perspectives of the approach.

Resume : Recently there has been enormous progress in the production of atomic layers that are strictly two-dimensional (2-D) and can be viewed as individual planes of atomic-scale thickness pulled out of bulk materials. Silicon is a leading material in semiconductor industries because of its wide usage in transistors, solar cells, and other electronic devices. Silicene, the silicon equivalent of graphene, has recently attracted significant attention. In this presentation, first I will describe our recent molecular dynamics simulation on thermal transport of atomically thin silicene with improved Stillinger-Weber potential. We found that the silicene has intrinsically extremely low thermal conductivity (~ 10 W/mK at 300 K), which is about 15 times smaller than bulk Si and two orders of magnitude less than that of its carbon counterpart graphene. By quantifying the relative contribution from different phonon polarizations, we show that the phonon transport in silicene is dominated by the longitudinal modes, not the out-of-plane flexural modes as opposed to graphene. Second, I will demonstrate the effect of size (length) and surface functionalization on the thermal transport of silicene. We discovered that, contrary to its counterpart of graphene and despite the similarity of their honeycomb lattice structure, silicene exhibits an anomalous thermal behavior upon surface functionalization: the thermal conductivity of silicene significantly increases with applied hydrogen percentage. Third, by conducting a series of non-equilibrium molecular dynamics simulations, we show that substrates have great effect on the thermal conductivity of silicene. More importantly, substrates can tune the thermal conductivity of silicene in a broad range, which is very important for the thermal management of electronic devices involving silicene. Our findings provide a guide of how to modulate the thermal transport properties of two-dimensional Si with nanoengineering and may be of use in tuning their electronic and optical properties for electronic, thermoelectric, photovoltaic, and opto-electronic applications.

Resume : There are two main approaches currently used for analysing phonon thermal transport in crystals. The first approach employs the Green-Kubo formalism which basically requires the use of molecular dynamics simulation for an equilibrium calculation of the heat current autocorrelation function (HCACF). The second approach is based on employing the Boltzmann transport equation for the phonon distribution function. The first approach is microscopic as it deals with microscopic definition of the heat current. It needs only an appropriate interatomic potential as input and does not require any assumptions about the nature of thermal transport in the model system. Meanwhile, the second approach is used for phenomenological interpretation of experimental and simulation data on phonon thermal transport. In this contribution, we discuss Green-Kubo calculations of the phonon thermal conductivity of crystals with a monatomic unit cell, which revealed: (i) two-stage decay in the HCACF; and (ii) deviation from the inverse temperature dependence of the phonon thermal conductivity at high temperatures towards a more rapid decrease following an exponent close to -1.4. Then we use a general expression for the phonon thermal conductivity of an isotropic solid which follows from the Boltzmann transport equation in order to facilitate analytical treatment of the results of the Green-Kubo calculations.

Resume : We present here an exhaustive ab initio study of the use of carbon-based tips as electrodes in single molecule junctions. Motivated by recent experiments, we show that carbon tips can be combined with other carbon nanostructures, such as graphene, to form all-carbon molecular junctions with mol?cules like benzene or C60. Our results show that the use of carbon tips can lead to relatively conductive molecular junctions. However, contrary to junctions formed with standard metals, the conductance traces recorded during the formation of the all-carbon single-molecule junctions do not exhibit clear conductance plateaus, which can be attributed to the inability of the hydrogenated carbon tips to form chemical bonds with the organic molecules. Additionally, we explore here the use of carbon tips for scanning tunneling microscopy and show that they are well suited for obtaining sample images with atomic resolution.

Resume : The way ceramic nanoparticles behave has recently gained more interest, especially in the field of modern surgery where metallic alloys used for implants and prosthesis are progressively replaced by bio-compatible ceramics to increase their lifetime and reduce ion release responsible for inflammatory reactions. As bio-compatible ceramics are originally made of powders, one may wonder whether size effects could play a role during compaction and sintering and thus, influence the post-processed material mechanical properties. Based on this framework, this study investigates the mechanical behaviour of isolated ceramic nanoparticles using statics and molecular dynamics. Here we model magnesium oxyde MgO as it is a model ceramic, widely studied in terms of elastic properties, slip systems, dislocation characterization and flow behaviour under compression at the macroscopic scale. The ionic system is described using a Buckingham potential and a pairwise Coulomb interaction between point charges. Numerical nano-mechanical tests show that MgO nanoparticles deform up to large strain in comparison to bulk material, without any sign of damage. Deformation proceeds by ½<110>{110} dislocation nucleation and multiplication. Results are interpreted within the framework of small-scale plasticity. Finally, these findings enable the interpretation of recent in situ TEM compression tests performed on MgO <100>-oriented nanocubes at scale about 100 nm.