Calculation of the Properties of Chemically Active Nonequilibrium Plasma

From atoms and molecules to new materials and technologies

Predictive Modeling of Advance Materials and Material Processing Based on Multiscale Simulation Paradigms

Boris Potapkin Kintech Laboratory Ltd

Presented at International School on “Computer simulation of advanced materials” MSU, Moscow July 2012 Approach

Standard approach to design of new materials and technologies: empirical search

Material, Idea Prototype Testing technology

ü Expensive ü Long time ü No assimilation of fundamental data

The role of modeling consists in description and extrapolation of experimental data using phenomenological models

Integrated approach, based on predictive modeling

Material, Idea Modeling Prototype Testing technology

ü Reduced time and cost of development ü Reduced risks

A priory modeling, based on detail understanding of structures and mechanisms at atomistic level, before construction of prototype Technologies and tools of predictive modeling Problems ü Huge computers resources needed even for every single level EngineeringEngineering design,design, ü Unfeasibility of direct integration of spatial and temporal scales UnitUnitprocessprocess design design Direct integration ü No reliable integration methods FiniteFinite element element Analysis,Analysis, ü Stochastic nature of multilevel modeling: models, data, properties ContinuumContinuum modes modes MesoscaleMesoscale ü Cognitive problems ( data formats, computer languages etc.) modeling,modeling, MCMC Molecular Molecular Mechanism ü No effective software tools for scales integration and collaborative dynamicsdynamics (MD) work QuantumQuantum (MD) mechanicsmechanics (QC)(QC)

Background for solution ü Exponential growth of computer resources (Moore’s law, HPC) ü Development of high-performance algorithms ü Development of predictive theoretical methods Solution Development of methods and tools for integrated multilevel modeling

Elaboration and use of High Performance Computing algorithms and systems Multiscale Modeling for Advanced Materials

Fuel Cells, Energy Electrochemical Organic Matrix Materials batteries Nano composites Metamaterials Focus on: Catalysts Materials Properties OLED, LED, PV, OPV Simulation Paradigms

Macro level : FE, FDTD for Mesoscale methods: DPD, Mean Atomistic methods: quantum optics Field Models, MC chemistry, molecular dynamics, DFT

High Performance Computing Code parallelization HPC Architecture choice Platform-specific Code Optimization (Tailoring)

Information Technologies for Cloud Computing & Distributed Collaboration Kintech Lab profile

KINTECH was founded in 1998 by scientists and engineers from and the NRC "Kurchatov Institute“, MIFI, and Moscow State University

ACTIVITY FIELDS: ü Conducting of inventive research and consulting for a wide range of applications ü Software development for multi-scale multi-physics modeling KINTECH Lab, Ltd. modeling and design in complete cycle 1, Kurchatov Sq., Moscow, ü Customer support in their own research activity using the Russian Federation, advanced simulation capabilities of KINTECH's software (+7 (499) 196-78-37 7 +7 (499) 196-99-92 - [email protected] www.kintechlab.com Selected Kintech projects in energy and materials: materials & devices

ü Multiscale modeling of thin high-k dielectric film deposition and the investigation of their properties ü First-principles modeling of defects at the SiC/SiO2 interface ü Multiscale modeling of the optic properties of metamaterials and the design of devices on their basis ü Multiscale modeling of semiconductor nanosensors and CdTe solar cells ü Multiscale modeling and screening of scintillator and phosphor materials ü Optimization of the fabrication of microelectromechanical (MEMS) devices ü Multiscale modeling of phase transition in ferroelectric materials ü Modeling of Nano Electro Mechanical systems (NEMS) based on carbon nanotubes ü First-principles modeling of catalysts for fuel cells ü Predictive modeling of carbon based materials Selected Kintech projects:projects in energy and materials :

ü Multiscale modeling for PDE design ü Multiscale modeling of cold spay technology ü Multi-physics modeling of car exhaust cleaning ü Mechanistic modeling of depleted combustion processes ü Mechanistic modeling of coal gasification kinetic ü Plasma waste gasification modeling ü Industrial safety: explosions ü Multiscale modeling of chemically active plasma systems including PAC and plasma exhaust cleaning modeling ü Multiscale modeling of chemically active plasma systems ü Membrane gas separation modeling and system design ü First-principles modeling of catalysts for fuel cells ü Software development for environmental and industrial safety Kintech selected customers

ü General Electric since 2003 ü Motorola (Freescale Semiconductor), 2000-2006 ü Intel since 2007 ü Siemens ü Daimler ü Qualcomm ü David System and Technology SL, Spain ü Scientific Utilization Inc. USA ü Renault, France ü Rhone Poulence, France ü TNO, Netherlands ü DLR (Germane Airspace Center), Germany ü Arvin Meritor Inc, USA. ü PlasmaSol Corp., USA ü Princeton University, USA Key Parallel Software tools Developed at Kintech

Kintech Lab develops methods and special software tools for multilevel modeling in different engineering fields:

üChemical Workbench – an integrated environment for the development and reduction of chemical mechanisms for combustion, plasma, etching, films growth üFDTD-II – a tool for modeling the optical properties of metamaterials üMD-kMC – an integrated environment for atomistic modeling üEtchLab – a tool for modeling and optimization of MEMS fabrication üTRACC - integrated package to solve 3D fluid dynamics problems with radiation transport using special software and a database Predictive multi scale modeling for energy and materials: specific projects

KINTECH Lab, Ltd. 1, Kurchatov Sq., Moscow, Russian Federation, (+7 (499) 196-78-37 Advanced plasma light 7 +7 (499) 196-99-92 - [email protected] source www.kintechlab.com Theoretical Screening of Effective Emitting Substances in Fluorescent Plasma Light Sources

Goal: mercury free light source development

Emitting systems

Number of metals - 54 Ligands number - 8 Problems Large number of candidates Number of system to explore: = 486 Advanced plasma sources of light

Hg-free lamps on metal halides non-equilibrium plasma

I (-) +M*=>I + e +M GaI3 + e =>GaI3(-)=>…. 2 2 Exp. GE spectrum =>…GaI+e.=>Ga + I,I2(-) j j Ga +e=>Ga*=> Ga + hw Рисунок лампы с травлением

evaporation etching condensation

Ga, GaI2, GaI3 (wall) GaI3(pellet)

Ar : GaI3 New corrected set P (Ar cold) = 2 Torr, R = 1.27 cm,I = 300 mA 6 tube 1 1.0 3 T = 79 C ( [GaI ] = 3.3 1013 cm-3 ) GaI( p) GaI Dissociation by electron impact 3 through different dissociable terms 2 2 0.8 Ga( P)+I( P) 2 4 cm 0.6 GaI(A) -16 theory

0.4 2 b 10 Q(u), kXA AX 0.2

GaI(X) 0.0 Emission power, mW/cm 0 X 0 10 20 30 40 50 200 250 300 350 400 450 u, eV Wavelength (nm) Non-empirical multi physics multi scale approach to calculation of the properties of chemically active nonequilibrium plasma

Plasma modeling for parameters Recommendations optimization and processes control

Construction and analysis of a physical - chemical mechanism

Ga(4p1/2)+GaI3=>GaI+GaI2 Ga(4p1/2)=>Ga(Wall) GaI=>GaI(Wall) GaI2=>GaI2(Wall) Evaluation of cross I=>I(Wall) Ga(4p3/2)=>Ga(Wall) Ga(4p1/2)+e=>Ga(5s)+e Ga(4p3/2)+e=>Ga(5s)+e

sections, rate constants, Ga(4p1/2)+e=>Ga(4d)+e Ga(4p3/2)+e=>Ga(4d)+e GaI+e=>Ga(4p1/2)+I+e and radiation parameters GaI2+e=>GaI+I+e GaI3+e=>GaI+I2+e GaI+e=>Ga(4p3/2)+I+e -0.6 -0.4 -0.2 0.0 0.2 0.4 Sensitivity ab initio calculations of unknown parameters of atoms, molecules, and their interactions DATA ACQUISITION BY AB INITIO ELECTRONIC STRUCTURE CALCULATIONS

During the last decades computational electronic structure methods have become Ø competitive to experimental techniques in the accuracy of excited state energies (0.1 eV and better) Ø often more accurate than experiment in transition probabilities, oscillator strengths Ø reliable source of information on the states & transitions difficult for experimental studies

Highly accurate but expensive ab initio methods (coupled-cluster like)

Intermediate accuracy approaches (e.g. multireference perturbation theories)

Roughly approximate methods (e.g. time-dependent DFT) EXCITED ELECTRONIC STATES OF MOLECULES

Highly accurate but expensive ab initio methods

Approximate MultiReference Coupled Clusters = size- Exhaustive info on consistence corrected Small (2-3 non-H low-lying states of configuration interaction atoms) molecules small light-element (MR AQCC, ACPF) molecules Columbus, Asec 3

Excitation energies: Response / Green function < 0.1 eV errors techniques Dalton, Gaussian, Molpro Molecules with Cfour, Asec 3 simple shell Oscillator structure (closed- strengths: shell or one open Relativistic ± 10-20 % and shell) formulations for Fock-space coupled cluster better heavy-element methods compounds Asec 3 available EXCITED ELECTRONIC STATES OF MOLECULES

Intermediate accuracy approaches

Small & medium size Multiconfiguration molecules Scanning of perturbation theories (~101 non-H atoms) potential surfaces (PT): CASPT2 etc (fails in certain Molpro, Molcas areas) Firefly, Games Excitation energies: ~ 0.1 eV accuracy Multireference (effective-Hamiltonian) many-body Scanning of Perturbation Theory Oscillator strengths potential surfaces Relativistic (PT): ± 10-30 % (MCQDPT, & transition formulation of quasidegenerate PT MPPT) moments for states MPPT exists (MCQDPT) of any nature multipartitioning PT Etc Efop EXCITED ELECTRONIC STATES OF MOLECULES

Roughly approximate methods

Time-dependent DFT Medium size (MCSCF) Gaussian, ADF, etc and large (~101 non-H atoms) molecules

Main way to study Excitation energies: excitations for large ~1 eV accuracy molecules; Multiconfiguratinal Self (TD DTF – often much useful for geometry Consistent Field (MCSCF) better) optimization of excited Gaussian, Molcas, Molpro states

Oscillator strengths: Time-dependent Hartree- qualitative Fock (TDHF) CI / singles (CIS) All the codes Theory & modeling highlights for electronically excited states

Accurate (< 0.1 eV ) first-principles calculations of electronically excited states are feasible!

Example excited states and transition dipoles of InI

Multiple avoided crossings High states densities

Dependence on R is non-trivial ! GROUND ELECTRONIC STATES OF MOLECULES

Highly accurate but expensive ab initio methods

Multireference singles + Transition state (TS) of doubles configuration chemical reactions interaction (MR SDCI) Small & medium size (term crossing) method molecules Scalar relativity (about 10 atoms) through relativistic Fourth-order Moeller- Accurate calculations effective core Plesset (MP4) level of of potential energy potentials (RECPs) perturbation theory Barrier heights and surface (PES) only in for heavy element heats of reactions: the vicinity of its compounds Coupled Cluster accuracy of stationary points Not for the right calculations (CCSD, 0.1 eV Too expensive for lower corner of CCSDT, etc) surface scan! periodic table. Gaussian, Molpro, etc

Barrier heights and Highly accurate Quasi Additive heats of reactions with estimates for energies Compound Methods accuracy of 0.1 eV for of molecular processes (CBS - - Complete Basis molecules containing by combining several Set Extrapolation up to 20 atoms different approaches. methods; The cheapest ! G1, G2, G3 - Gaussian - To be verified first ! 1, 2, 3 methods etc) Gaussian GROUND ELECTRONIC STATES OF MOLECULES

Intermediate accuracy approaches

Second-order Moeller-Plesset Medium size Scanning of PES & (MP2) perturbation molecules evaluation of various theory (10-100) properties of All the programs molecules and Spin-orbit DFT Turbomol chemical reactions method taking into accuracy of account relativistic

0.2 eV – 0.3 eV and effects for adequate

can be less description of

For the Ga and In molecular parameters DFT Methods Medium size systems MP2 was for heavy element (GGA and hybrid & large ( hundreds of proved to give 0.1 eV compounds functionals) atoms) molecules accuracy All the programs

Accelrys

ACCURACY OF AB INITIO CALCULATED PROPERTIES OF MOLECULES AND CHEMICAL REACTIONS

Molecular property Accuracy Light element compounds Heavy element (neglect the effects of compounds relativity)

Equilibrium interatomic 0.01 Å 0.01-0.05 Å distances

Bond angles 1 º 1 – 3 º

Vibrational frequencies 1 – 3 % 5 – 10 %

Dipole moments 0.1 D 0.2 – 0.5 D

Polarizabilities 10 % 10 – 30 %

Barrier heights and heats 1 - 2 Kcal/mol 5 – 10 Kcal/mol of reactions

First ionization potentials and 0.1 eV 0.2 – 0.5 eV electron affinities EXAMPLE: ENTHALPY OF REACTION 2AlCl2 -> AlCl + AlCl3

0 kcal/mol

-18 kcal/mol-18.0 (old experimental value, NIST)

-42 kcal/mol (our ab initio estimate)

• AlCl2 enthalpy of formation value was revised (-67.0 to -57.1 kcal/mol) • New experimental estimate for reaction enthalpy -37 kcal/mol Calculation of reaction rates coefficient and kinetic mechanism build up

Reactions rate coefficients calculation Mechanism development

-10 3 k = 7´10 cm /s

– – Reaction profile GaI2+ GaI3 ® GaI + GaI4 Non-empirical approach to calculation of the properties of chemically active nonequilibrium plasma: validation Atomic Emission Ar-GaI3 System Molecular Emission Ar – GaI3 System Glow Discharge Glow Discharge 6 5s => 4p3/2 5s 5s => 4p1/2 5s 5 Simulation 1.0 Simulation Experiment 6s => 4p3/2, 1/2 => 4p3/2, 6s Experiment 4 0.8 4d => 4p1/2 4d => 4p3/2 3 0.6

2 0.4 Relative intensity

Emission Intesity, a.u. 1 0.2

0 0.0 260 280 300 320 340 360 380 400 420 380 384 388 392 396 400 Wave length, nm Wave length, nm

Predictive system models & understanding can be built up from first-principles estimates of underlying physical & chemical kinetics Sensitivity analysis and parameters optimization

Sensitivity analysis: Discharge parameters optimization why calculation we can be predictive

Ga(4p1/2)+GaI3=>GaI+GaI2 Ga(4p1/2)=>Ga(Wall) 40 GaI=>GaI(Wall) GaI2=>GaI2(Wall) 30 I=>I(Wall) Ga(4p3/2)=>Ga(Wall) Ga(4p1/2)+e=>Ga(5s)+e 20 Ga(4p3/2)+e=>Ga(5s)+e

Ga(4p1/2)+e=>Ga(4d)+e 10 Ga(4p3/2)+e=>Ga(4d)+e

GaI+e=>Ga(4p1/2)+I+e Efficiency, Emission % GaI2+e=>GaI+I+e 0 4 GaI3+e=>GaI+I2+e 8 Ar Pressure, Torr 90 100 GaI+e=>Ga(4p3/2)+I+e 12 80 16 70 -0.6 -0.4 -0.2 0.0 0.2 0.4 20 60 Sensitivity Temperature, C Predictive multi scale modeling based on HPC for energy and materials: specific projects

KINTECH Lab, Ltd. 1, Kurchatov Sq., Moscow, Russian Federation, (+7 (499) 196-78-37 Theoretical screening of 7 +7 (499) 196-99-92 - [email protected] advanced phosphors www.kintechlab.com Modeling of phosphor properties

Optimization of conversion efficiency of phosphors

Search and selection of best phosphors with prescribed properties

Large Stockes shift of 5d states in the 3+ 3+ LaPO4:RE and YF3:RE phosphors is related to considerable changes in the coordination number of the rare-earth site upon the 4f ®5d excitation Modeling of phosphor properties

Термодинамический анализ материала Расчет спектров поглощения и испускания, интенсивности и других свойств люминофора

30 Расчет миграции и

20 захвата

10 возбуждений

0 -25 -20 -15 -10 -5 0 5 10 15 Energy, eV Первопринципный расчет электронной структуры материала

Восстановление LuCl3 LuBr3 LuI3 структуры материала из дифракционных данных Modeling of phosphor properties

ligands Coordination number V = v(k) of Ce in LaРO CF å 4 k changes from 9 to 8

Calculated Stocks shifts

Host/dopant Calc (cm-1) Exp (cm-1)

3+ LaPO4:Ce 5050 4884 3+ LaPO4:Pr 4300 - 3+ YF3:Ce 5100 5444 3+ GdF3:Ce 5273 5567 3+ K2GdF5:Ce 2005 2227 Predictive multi scale modeling based on HPC for energy and materials: specific projects

KINTECH Lab, Ltd. 1, Kurchatov Sq., Moscow, Russian Federation, (+7 (499) 196-78-37 A priory design of Photonic 7 +7 (499) 196-99-92 - [email protected] Metamaterials www.kintechlab.com Photonic metamaterials

Properties of photonic metamaterials are defined by: vchemical composition

- 1000 mkm vmicro- and nano-scale geometry !

L ~L 10 Feature size I ~ 100 nm Applications: v New light sources emitting in controlled spectral region v Emission control for creation of new luminescent materials and device Photonic metamaterials: v Optoelectronics design of new nonlinear optical materials vModify radiation/reflection pattern v Creation of new lasing media, photonic v Enhance reflection in forbidden bands fiber laser vEnhance emission in allowed (transmission) v Mirrors, photonic waveguides, couplers bands and multiplexers vReduce radiation in band gaps v Beam shaping, new types of fibers vModify density of states v New elements for near field optics

(L/l)3 >> 106 : Parallel Code Development & HPC are of critical importance ! Parallel FDTD code for metamaterials design Finite differences time-domain method (FDTD) & code üSimulation of arbitrary geometry üSimulation of materials with nonlinear material properties üSimulation of E-M field distribution inside and outside the structure üSimulation of oblique incidence on periodic structures üSensitivity analysis (modeling the impact of defects)

üNew light sources emitting in controlled spectral region üEmission control for creation of new luminescent materials üOptoelectronics design of new nonlinear optical materials Applications üCreation of new lasing media, photonic fiber laser üMirrors, photonic waveguides, couplers and multiplexers üBeam shaping, new type of fibers üNew elements for near field optics üCreation of photonic materials with electrically or mechanically controlled characteristics üNew left handed materials Electromagnetic Template FDTD Library underlining FDTD code

üNovel computational methods within FDTD in EMTL: Ø Subpixel smoothing for dispersive media, for reducing the staircasing affects of the media interfaces on a regular grid Ø Iterative technique for simulation of oblique plane wave incidence on a periodic structure Ø The method of calculation of the frequency transfer matrix by FDTD for simulation of optical properties of multi-layered periodic structures üEMTL is a parallel library (MPI, Open MP) üCross platform üHigh computational efficiency for arbitrary problem geometries is achieved by balanced domain decomposition üLinear parallel scalability even for large numbers of processors

• A. Deinega and I. Valuev, Optics Letters 32, 3429 (2007) • I. Valuev, A. Deinega, and S. Belousov, Optics Letters 33, 1491 (2008) • A. Deinega, S. Belousov, and I. Valuev, Optics Letters 34, 860 (2009) • A. Deinega and I. Valuev, Computer Physics Communications, to be published (2010) Numerical methods and programs for optical properties modeling

2. Layered Korringa-Kohn-Rostoker method (LKKR): 3. Plane wave expansion method (PW): for spectra calculations of ideal photonic crystals with + for band structure of ideal photonic finite width crystals calculations (1D, 2D and 3D + for band structure of ideal photonic crystals symmetries) calculations + for density of photonic states and local + for density of photonic states calculations of ideal density of states calculations of ideal photonic crystals photonic crystals + taking into account experimental dielectric function + arbitrary shape of scatterers (real and imaginary part) for any material + high speed and convergence of the method for scatterers with spherical symmetry 5. Effective media method + for modeling of light propagation through 4. Ray-tracing method the random inhomogeneous medium + for modeling of light propagation through the with small impurities (much smaller than medium structured on the big scale (much the wavelength). Effective refractive bigger than the wavelength). The numerical index instead of complex structured realization of geometrical optics case. medium. Predictive multi scale modeling based on HPC for energy and materials: Advanced Light Source

Motivation

Metallic photonic crystals offer a fine visible emission control – a way towards high efficiency light sources

IR intensity IR intensity wavelength wavelength For maximizing efficiency of a light source optimizing photonic crystal material and Problem geometry is of crucial importance. A priori modeling greatly reduces development time to address and costs. Modeling of photonic crystal optical properties combined with the first principle based modeling of material properties solves the problem.

Solution Modeling and optimization of Kintech Lab and photonic crystal optical GE Global Research properties developed parallel FDTD code for CAD of materials and devices for nano-photonics and combined it with DFT First principle computation method for material of material properties properties modeling. Calculation of material optical properties at high temperature

Modification of electron density function Calculation method (DFT) (E. Maximov, UFN, 170, 1035 (2000))

Model takes into account: Ø Thermal expansion of solid bodies Ø Variations of electron occupation numbers with temperature Ø Interaction of electrons with ü Lattice thermal vibration ü Defects

1. Strong dependence of luminosity on the temperature 2. Good agreement with experimental data at S. Roberts, Phys. Rev. 114, 104 (1959) high temperatures Experiment Computer aided design of advanced light sources based on photonic crystals 2 мкм First principle computation of W material properties

SiO2 Modeling and optimization of 200 nm photonic crystal optical properties Taking into account defects of N=2, f=0.03periodic structure 0,9

0,8 1. Comparison to expt. data

0,7 2. Design rules 0,6 N=1, а=550нм, d=200нм (T=298K) a=400нм 0,6 N=1, а=550нм, d=200нм эксперимент 0,3 0,5 a=550нм 0,5 расчет: монослой W опала эксперимент идеальный кристалл (расчет) 0,4 a=700нм 0,25 0,4 разупорядоченный кристалл (расчет) 0,3 0,2

0,2 0,3 коэффициент поглощения коэффициент 0,15 0,1 h 0,2 W 0 0,1 dW 0,2 0,4 0,6 0,8 1 1,2 длинакоэффициент поглощения 0,1 волны , мкм hpost 0,05 коэффициентпоглощения 0 dpost 0 0,2 0,4 0,6 0,8 1 1,2 1,4 0,2 0,4 0,6 0,8 1 1,2 1,4 длина волны, мкм длина волны, мкм Predictive multi scale modeling based on HPC for energy and materials: Antireflection Coating, OLED outcoupling KINTECH Lab, Ltd. 1, Kurchatov Sq., Moscow, Russian Federation, (+7 (499) 196-78-37 A priory design of Photonic 7 +7 (499) 196-99-92 - [email protected] Metamaterials www.kintechlab.com Numerical modeling and optimization of parameters for antireflection coating based on periodic surface nanostructured for solar batteries application

Y. Kanamori, M. Sasaki, and K. Hane, Opt. Lett. 24, 1422 (1999)

Zhaoning Yu er al., J. Vac. Sci. Technol. B 21(6), 2874 (2004) Modeling OLED outcoupling efficiency enhancement with 2D patterned PC structures

Planar OLED PC OLED Glass substrate (n=1.5) SiNx 600 nm (n=1.90) ITO 200nm (n=1.80

Organic (n=1.75) HTL 100nm OLED structuresOLED ETL metallic cathode 80nm Outcoupling 1 1 1 1 0.8 0.9 0.8 0.9 efficiency and 0.6 0.8 0.6 0.8 directionality 0.4 0.7 0.4 0.7 0.2 0.6 0.2 0.6 improvement! 0 0.5 0 0.5 -0.2 0.4 -0.2 0.4 -0.4 0.3 -0.4 0.3 -0.6 0.2 -0.6 0.2 field angular angular field distibutions -0.8 0.1 -0.8 0.1 -1 -1 0 -1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1 Far -

Ref: Y.-J. Lee et.al. “A high-extraction-efficiency nanopatterned organic light-emitting 41 diode”, Appl. Phys. Lett. 82, 3779 (2003) Modeling OLED outcoupling efficiency enhancement with 1D cathode grating structures

Up to 2-3 times outcoupling efficiency enhancement

0.7 2.2 no grating regular grating: a = 0.4um, d = 0.1um 2 0.6 1.8

0.5 1.6

1.4 0.4 1.2 extraction efficiency enhancement factor 0.3 1

0.2 0.8 0.4 0.45 0.5 0.55 0.6 0.65 0.4 0.45 0.5 0.55 0.6 0.65 wavelength, mm wavelength, mm Predictive multi scale modeling based on HPC for energy and materials: Optical Chemical Sensors

v Luminescence suppression and amplification on certain wavelengths with photonic band gap (PBG) and local density of state (LDOS) maxima in photonic crystals v Local field enhancement in sensor layer v Overall emission intensity enhancement due to effectively large surface area of a sensor layer v Redistribution of the emitted light through the angles due to the scattering within nanostructured sensor layer v Efficiency managing of the energy transfer between donor-molecule and acceptor- molecule v Light entanglement effect for the absorption enhancement v Photonic crystal based antireflection coating usage for pump radiation enhancement

• W. Zhang et.al., Sensors and Actuators B: Chemical Vol.131, N1, p.279 (2008) • H.J. Kim et.al., Sensors and Actuators B: Chemical Vol.124, N1, p.147 (2007) • Z. Yang et.al., Optics Letters, Vol. 33, 17, pp. 1963-1965 • B. Kolaric et al, Chem. Mater. 19, 5547 (2007) • K. Shibata et al, Colloid Polym. Sci. 285, 127 (2006) • Carl Hägglund et al, Appl. Phys.Let. 92, 013113 (2008) • L. Tsakalakos et al, Appl. Phys. Lett. 91, 233117 (2007) Chemical sensor: multiscale approach based on Ab initio Q-chemistry and electro-magnetic predictive modeling

Optical response of the sensor layer calculation

FDTD modeling of electromagnetic field distribution in sensor layer and calculation of supramolecular centers radiation within photonic crystal

Absorption lines shape for definilaminoacredine complexes with acetone and benzol 180 Dipole in PC 160 Dipole in free space 140 Computation of absorption and 120 100 fluorescence lines shape of 80 60 Intensity, arb. units arb. Intensity, molecular complexes 40 20 0 0.4 0.5 0.6 0.7 0.8 0.9 1 Wavelength, mkm TDDFT computation of absorption and fluorescence Fluorescence intensity of free molecule in lines of supramolecular centers vacuum and in photonic crystal Chemical sensors: numerical modeling of the optical response

Model parameters: I(l,q,f ) Dye molecules ØStructure geometry: on the Ølattice type nanoparticle Ø size of elements surface Ønumber of layers PC layer ØMaterial properties e(w): Øelements optical properties Substrate Ømedium optical properties Øsubstrate optical properties ØDye molecules spectra ~ 200-1000 nm Ø absorption Pump Ø luminescence

Excitation of supramolecular Luminescence in Setting Pump wave complexes proportional to photonic crystal parameters propagation the local field intensity structure Numerical modeling of the optical response in optical chemical sensors Response Idye+analyte enhanced >10 times enhanced

Idye suppressed

PC parameters: a = 279 nm, h = w = 69.77 nm

Log-piles: enhancing dye+analyte peak by band-edge resonance, suppressing free dye peak by PBG Numerical modeling of the optical response in optical chemical sensors Response e =12.96 Idye+analyte enhanced > 5 times enhanced

Idye partly suppressed

PC parameters: a = 416nm, rsph= 88 nm (diamond close-packed case, f = 34%) Diamond PC: enhancing dye+analyte peak, suppressing free dye peak by PBG Predictive multi scale modeling based on HPC for energy and materials: specific projects

KINTECH Lab, Ltd. 1, Kurchatov Sq., Moscow, Russian Federation, (+7 (499) 196-78-37 Investigation of thermal 7 +7 (499) 196-99-92 - [email protected] conductivity of graphene www.kintechlab.com Thermal conductivity of graphene

Thermal conductivity of supported graphene

The extremely high thermal conductivity in the range of 3080– 5150 W/m K and phonon mean free path of 775 nm near room temperature. Exceeds graphite and CNT thermal conductivity (2000-3000 W/mK) Experiment: thermal conductivity depends on the number of graphene layers Theory: thermal conductivity of graphene increases with length

S. Ghosh et al, APL 92, 151911 (2008)

Mechanistic understanding and modeling prove is required to “believe” and to emply the effect for real graphene-based materials (e.g. graphene paper) Thermal conductivity of graphene/graphite 1 v Boltzmann Transport k = åò C(w,T ) ×t (w,T ) × ×w × dw Equation theory 4p × h l u Three-phonon scattering relaxation time: 1 kT w 2 P. G. Klemens, Journal of Wide = g 2 Bandgap Materials7, 332 (2000) 4 2 t Mv wD at w ® 0 w ® t w ® w 2 k ® w w = w ® ¥ C( ) kB , ( ) 1/ , ò d / ln( ) graphite graphene Minimum cutoff frequency is determined by Phonon mean free path is restricted by inter-plane interaction flake size: 1 M × v 2 w l(w,T ) = × v × × D < L 2 ×g 2 k ×T w 2

2 1 M ×v wD Þ wmin = ×v × × ZO’ mode 2×g 2 k ×T L

Þ Strong mode Graphene thermal conductivity should coupling increase with flake size Marzari et al, PRB 71, 205214 (2005) wmin = wZO' » 2p ×4THz Thermal conductivity of graphene/graphite Non-equilibrium MD modeling by Kintech parallel MD-kMC code

Q l = - bdÑT

Linear scaling was proven up to 200 cores • High-performance molecular dynamics simulations

should be used to model transport in real scale 60

graphene flakes: 50

• Flake sizes: micron scale 40

• Number of atoms: > 100,000 30 Время, с Время, efficient parallel MD algorithms are required for 20

many-body interatomic potentials (Tersoff, Brenner). 10 5 One week run on 200 cores for 10 atoms. 0 0 200 400 600 800 1000 1200 1400 • Domain decomposition method was adapted for n NEMD calculations Thermal conductivity of graphene/graphite

Boltzmann transport equation (BTE) 6000 Experimental Multi-scale modeling of 5000 graphene (BTE) values thermal conductivity: graphite (BTE) 4000

3000

Non-equilibrium MD (NEMD) modeling 2000

1000 Thermal conductivity, W/mK

900 graphene (NEMD) 1000 graphite (NEMD) 800 0 700 0.1 1 10 100 1000 600 Flake length, mkm

500 MD and BTE modeling prove that 400

300 1. Graphene thermal conductivity Thermal conductivity, W/mK conductivity, Thermal

200 increases with flake size

100 2. Interplane coupling in graphite 0 limits thermal conductivity 0 0.2 0.4 0.6 0.8 1 Flake length, mkm Thermal conductivity of graphene/graphite

Influence of defects on thermal conductivity of graphene

1000 nm with vacancies 1000 nm with OH groups

1800 1 600 1600 NEMD NEMD 1 400 1400 analytic Bose 1 200 BTE 1200 analyticBTE Bose BTE 1000 1 000 analytic classic analytic classic 800 BTE 800 W/mK 600 600 400 400 200

200 conductivity, Thermal

Thermal conductivity, W/mK conductivity, Thermal 0 0 0 1 2 3 4 5 0 1 2 3 4 5 Vacancy concentration, % OH groups concentration, %

Significant reduction of thermal Significant reduction of thermal conductivity at vacancy density about 1% conductivity at OH group density about 1% MD-kMC

MD-kMC code is an integrated environment for different atomistic simulations based on molecular mechanics, molecular dynamics, and kinetic Monte Carlo methods using a wide set of empirical and semiempirical energy functionals. Nonequilibrium molecular dynamics method implemented in MD-kMC allows to calculate phonon thermal conductivity.

MD-kMC library of potentials: • Charge variable potentials (QEq charge equilization method) • Environment-dependent potentials (Tersoff-type many body functionals (Tersoff, Brenner)) • (Modified) embedded atom (MEAM) potentials (Based on Baskes EAM and MEAM functionals) • Tight Binding methods for spd orbitals (Orthogonal TB, Orthogonal self- consistent charge (SCC) TB, Non-orthogonal TB, K-point sampling) http://www.kintechlab.com/products/md-kmc/ 55 Chemical Workbench

Thermodynamic Models – a set of models for calculating the thermodynamic properties of multicomponent mixture, Gas-Phase Kinetic Models – various models for general gas- phase kinetic modeling, Flame model – premixed Flame reactor is a 1D model for calculation of laminar flame front velocity and structure, Heterogeneous Kinetic Models – a set of models for surface chemistry modeling Non-Equilibrium Plasma Models – a set of comprehensive models for nonequilibrium plasma process, Detonation Model – model for estimating of wave parameters and modeling of advanced propulsion systems, Separators and Mixers – various tools to control reactor's streams, and a membrane reactor model for calculating separation characteristic of membrane unit, Mechanism Analysis and Reduction – a tool set for kinetic mechanism analysis and reduction. 56 http://www.kintechlab.com/products/chemical-workbench/ Khimera

Khimera allows one to calculate the kinetic parameters of elementary processes and thermodynamic and transport properties from the data on the molecular structures and properties obtained from quantum-chemical calculations or from an experiment. The molecular properties and the parameters of molecular interactions can be calculated using available quantum-chemical software (GAUSSIAN, GAMESS, JAGUAR, ADF) and directly inputted into Khimera in an automatic mode.

Khimera Models: Chemistry of Heavy Particles Surface Processes Electron-Molecular Reaction Vibrational Energy Transfer Photochemical Reactions and Electronic Energy Transfer Multicomponent Thermodynamic Properties Model Multicomponent Gas Transport Properties Model http://www.kintechlab.com/products/khimera/ 57 Drift-diffusion code (charge transport simulation) Solving Poisson and continuity equation for ì drift and diffusion of charge carriers ü Inorganic 1D and 2D heterojunction structures ïÑÑey + - + - = ü Steady-state solution ï qp( n NDA N)0 ï1 ¶n ü Shottky barrier or charge injection model at í Ñ+-=Jn GR =m +Ñ cathode/anode boundaries ¶ ïìJnn q nE qD n n ïqtí =m -Ñq Organic multilayer structures ï ¶ îïJpp q pE qD p p 1 p q Energy disorder ï-ÑJp + GR - = î qt¶ q Local mobility and diffusion coefficients

JV curves of cell with recombinative grain

Distribution of the recombination rate near boundaries

grain boundary in thin film solar sell 58 Predictive multi scale modeling based on HPC for energy and materials: specific projects

KINTECH Lab, Ltd. 1, Kurchatov Sq., Moscow, Russian Federation, (+7 (499) 196-78-37 A priory design of Combustion 7 +7 (499) 196-99-92 - [email protected] based Advanced Energy Systems www.kintechlab.com Development Cycle of Combustion Chamber Based on Predictive Modeling CFD Model FLUENT Integrated KINTECH tools Combustio n Mechanism Model Experiments Thermo- CWB Chemical Partial Modeling and Kinetic Data Ab initio Khimera Calculations QC Combustor detail Experiments design Shock tubes Flames, Flow Reactors etc Compression Machine Rapid Fundamental

Experiment Software for development and reduction of the combustion mechanisms of real fuel: Data recovery with Khimera®

GAMESS Gaussian

Jaguar ADF

Khimera® Kintech Lab & Motorola

Relevant First-principles based physico-chemical data: • Elementary processes in gas phase and plasma, at surface and in liquid phase • Gas mixtures transport properties • Thermodynamic data of individual substances Software for development and reduction of the combustion mechanisms of real fuel: Mechanism generation software

Khimera® Elementary processes (Kintech Lab)

H C C C H Generation and validation of detailed mechanisms for C C combustion in the frame of Chemical Workbench

Validated detailed kinetic mechanisms for reduction and CFD modeling Method for mechanism generation was developed by multi-disciplinary collaborative team supported by RFBR grant: RRC Kurchatov Institute (contractor), Semenov Institute of Chemical Physics RAS, Photo Chemistry Center RAS, Institute of Mechanics MSU, Kintech Lab Software for development and reduction of the combustion mechanisms of real fuel: HPC and mechanism reduction

4500 Detailed mechanisms for PRF (Curran et al.) 4000 real fuels combustion – 1000 3500 iso-octane exponential growth of (Curran et al.) Mechanism Reduction 800 3000 computational resources n-heptane Chemical Workbench® - parallel code for 2500 for simulation (Curran et al.) 600 kinetic mechanisms automatic reduction and 2000 multiple testing 1500 hydrogen 400 Number of Species Number of Reactions propane 1000 methane (Marinov) (GRIMech3.0) n-butane 200 Reduction methods 500 (ENSIC Nancy) • Computational singular perturbation 0 0 0 1 2 3 4 5 6 7 8 9 • Principal component analysis Carbon Number • Directed Relation Graphs • Rate-of-Production analysis

Code Parallelization Kintech Lab software Chemical Iso-octane mechanism reduction runs 43 Workbench® was parallelized and minutes instead of 24 hours with new optimized by Intel CRT team and parallelization algorithm. We have 2,27 Kintech Lab experts for Intel times performance boost of our application architecture with Intel Xeon 55xx vs. previous generation Intel Xeon 54xx. Predictive Modeling of New Generation Detonation Based Engine: Problem

Pulsed detonation engine (PDE) Surrogate of the aviation fuel at GE Global Research Centre

n-decane C10H22 - 72.7% n-hexane C6H14 - 9.1%

benzene C6H6 - 18.2%

phi=1. 1.E-01

1.E-02

1.E-03

1.E-04

1.E-05

Problem: s time, Induction 1.E-06 Develop predictive CFD model for simulation of 1000/T, 1/K 1.E-07 detonation initiation and propagation inside of the PDE, 0.6 0.7 0.8 0.9 1.0 which is capable to operate with standard aviation fuel – JetA in experiments at P=9.4 atm [1] JetA surrog. experiments. [1] aviation kerosene Jet-A Ignition delay time for aviation kerosene (GE GRC experiments) Predictive Modeling of New Generation Detonation Based Engine: Jet-A combustion mechanisms development

QC calculations for uncertain reactions Mechanisms validation

-20 3.2 C6H6+O2<=>C6H5/+HO2 k = 1.3∙10 T exp(-61.45/RT) Combustion mechanisms for Jet-A surrogate

1000 417 species reactions

100 71 37 23 10 11 10

1 detailed reduced global Detailed mechanism reduction Detailed mechanism (~ 400 rxn)

Skeletal mechanism (~ 100 rxn)

Reduced mechanism (~ 30 rxn)

Global mechanism (~ 10 rxn) Predictive Modeling of New Generation Detonation Based Engine: parallel computations load balance for CFD

Parallel computations performance – case study for detonation simulations • Unsteady detonation wave propagation • Global mechanism of Jet-A combustion (10 species, 11 reactions) • Intel Xeon CPUs volumetric initiation of detonation • Infiniband connection

1000

efficient parallelization For available cluster configuration data trasnfer between CPUs limits the calculation speed is was decided to limit the number

100 of cores by 36

Run time, hours Different tasks with 36 cores were run simultaneously for efficiently 10 use of computational resources 10 100 Cores Predictive Modeling of New Generation Detonation Based Engine: Jet-A fuelled PDE ignition modeling

Initiator Main chamber

PDE

Jet-A/О2 Jet-A/air Optimal composition? (stoichiometric)

Experiment fopt ≈ 0.4 f = 0.3 f = 0.4 f = 0.6 (detonation quenching) (wall-induced detonation (volumetric initiation of inittiation) detonation) Predictive Modeling of NOx emissions from industrial GT burner

Natural gas-fired burner for industrial gas turbine Chemistry models

Goal: develop predictive CFD model of NOx Detailed and Reduced mechanism for CH4 emissions from GT burner • 82 species and 191 reactions • 17 species and 25 reactions

Effects to be included due to complex mixture • reduced vs. detailed: maximum error in simulation of laminar flame velocity and ignition delay time 20% composition • complex chemistry of methane combustion NO formation mechanisms: • Radiative heat losses from burner to ambient • thermal mechanism • NOx formation paths • N2O path

ignition delay time: detailed vs. reduced,ER = 0.5, 1, P = 4 atm and 15 atm 1.E-01

1.E-02

1.E-03 Detailed ER = 0.5, P = 4 atm Reduced ER = 0.5, P = 4 atm tind, s tind, 1.E-04 Detailed ER = 0.5, P = 15 atm Reduced ER = 0.5, P = 15 atm 1.E-05 Detailed ER = 1, P = 15 atm Reduced ER = 1, P = 15 atm

1.E-06 0.0004 0.0006 0.0008 0.001 0.0012 1/T, 1/K Predictive Modeling of NOx emissions from industrial GT burner

3D mesh with ~ 106 cells

axial velocity

temperature

Average run time with reduced mechanism: • 18 Intel Xeon Dual Core CPU (36 cores) NO formation rate • Infiniband interconnect • 20 - 24 hour to reach steady-state solution

NO, ppm@15%O2 0.5*Full Power 0.8*Full Power Full Power Modeling 10 - 20 17 - 28 37 - 43 Experiment 23 27 33

Desired accuracy – 20% maximum error in NO concentration prediction – is attainable with reduced mechanisms of methane combustion Thank you!

KINTECH Lab, Ltd. 1, Kurchatov Sq., Moscow, Russian Federation, (+7 (499) 196-78-37 7 +7 (499) 196-99-92 - [email protected] www.kintechlab.com