Atomistic Modeling of nanostructured materials for energy and
Transcription
Atomistic Modeling of nanostructured materials for energy and
UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Atomistic Modeling of nanostructured materials for energy and biomedical applications Claudio Melis and Luciano Colombo Department of Physics, University of Cagliari (Italy) claudio.melis@dsf.unica.it Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Theory and simulations of nanomaterials@UNICA Research activity Novel (nano)materials for energy production harvesting, biomedical applications, and metrology Large-scale atomistic simulations aimed at the characterization of specific materials’s properties such as: morphological, electronic, thermal transport and mechanical Group members Prof. Luciano Colombo - Full Professor -luciano.colombo@dsf.unica.it Dr. Claudio Melis - assistant professor - claudio.melis@dsf.unica.it Dr. Konstanze Hahn - post-doc - konstanze.hahn@dsf.unica.it Ms. Giuliana Barbarino - Ph.D. student - giuliana.barbarino@dsf.unica.it Mr. Riccardo Dettori - Ph. D. student - riccardo.dettori@dsf.unica.it Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Current Research interests SiGe nanostructured alloys and superlattices for thermoelectric conversion; nanoporous Si for thermal insulation; graphene and graphane as “heat paths” for phononic devices; nanoengineered elastomers for next-generation deep brain stimulators; SiGe nanocomposites Graphene-based thermal diodes Claudio Melis Nanoporous Si Polymer/gold nanocomposites Atomistic Modeling of nanostructured materials for energy and biomedical applic Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Continuum models ……… µm Coarse grained models size UNICA Classical molecular dynamics nm Å First principles calculations ps fs ns …………. µs ms time Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Our role in the Freeeled project 1 Characterization of the thermal exchange between GaN/Ga( 1 − x)Inx N and the phosphor 2 Characterization of the geometrical/morphological features of the organic molecules as afunction of temperature heptazines heptazines GaN heptazines Ga1-xInxN Ga1-xInxN heptazines GaN UNICA d Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Example 1 Elastomers are subjects of an increasing interest for biomedical applications Poly-dimethylsiloxane (PDMS) couples biocompatibility with excellent elastic properties The final goal is to functionalize PDMS by integrating metallic circuits and microelectrodes PDMS-metal nanocomposites: Supersonic Cluster Beam Implantation (SCBI)* Implantation of neutral Au nanoparticles (Au-nc) inside a PDMS substrate Collimated beam of neutral metallic clusters towards a polymeric substrate NO charging or carbonization of the polymeric substrate * C. Ghisleri et al. J. Phys. D: Appl. Phys. 46 (2013) Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Open questions Atomistic characterization of the SCBI process 1 2 Cluster penetration depth vs. implantation energy Substrate morphology upon the cluster implantation Temperature Surface roughness Characterization of nanocomposite elastic properties Nanocomposite Young Modulus vs. metal volume concentration Atomistic simulations Computer simulations are a key tool to describe at the atomistic level: 1 2 SCBI process Nanocomposite elastic properties Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results SCBI simulations* *R. Cardia et al., J. Appl. Phys. 113, 224307 (2013) Computational setup PDMS chains: 100 monomers Au-nc 3-6 nm radius Impl. energ.: 0.5,1.0,2.0 eV/atom Classical MD, COMPASS force-field System size:∼5·106 atoms Cluster penetration depth vs implantation energy: linear dependence Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Substrate characterization upon the implantation* * C. Ghisleri et al. J. Phys. D: Appl. Phys. 46 (2013) Simulated surface topography map Temperature Craters on the PDMS surface Craters lateral dimension ∼ cluster dimensions Surface temperature increase Craters depth ∼ implantation energy Two temperature spots: Hot region (T∼ 320-350 K) Surface roughness increases with the implantation energy Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Elasticity of a polymer network Polymer deformation Assumptions 1 Volume conservation: λ1 λ2 λ3 = 1 2 Affine deformation 3 No change of internal energy upon deformation Deformation work 1 NKB T(λ21 + λ22 + λ23 − 3) 2 NKB T = G: Elastic modulus of the polymer network ∆S: Entropy variation upon deformation W = −T · ∆S = Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Elasticity of a polymer network Uniaxial deformation λ1 = λ √ λ2 = λ3 = 1/ λ W= 1 2 G(λ2 + − 3) 2 λ Stress-strain relationships Tensile stress σT True stress: force on the strained surface: σii = λi σ11 − ∂W − P0 ∂λi 1 σ22 + σ33 = σT = G(λ2 − ) 2 λ G is related to the Young modulus E: The constant P0 is due to the incompressibility of the polymer E = 3G Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Simulation procedure 1 2 3 4 Samples generation Uniaxial deformations Tensile stress sampling G estimation by linear interpolation Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Simulation procedure (1) Samples generation (a) Au-nc randomly distributed at 5 fixed concentrations ranging 5%-30% (b) NPT simulations for 2 ns: nanocomposite self-assembling United atoms force field : UBONDS + UANGLE + UDIHEDRAL + UCOUL + UVDW Experimental cluster size distribution Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Tensile stress calculation (2) Uniaxial deformation (4) G estimation: σT = G(λ2 − (3) σT sampling 10 % Au-nc, λ=1.2 Au-nc 10% 150 σΤ(Atm) σΤ(Atm) 200 100 50 0 0 1 2 3 Time (ns) 4 1 ) λ 5 Claudio Melis 60 50 40 30 20 10 0 0 0.5 1 1.5 λ2-1/λ Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Results Comparison with the AFM experiments Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Conclusions SCBI simulations The cluster penetration depth linearly depends on the implantation energy The substrate surface morphology is largely affected by the cluster impact PDMS-Au nc mechanical properties The PDMS-Au nanocomposite Young Modulus is unaffected up to ∼ 25% Au conentration Our results are in very good agreement with recent AFM nanoindentation experiments Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Introduction 1 Thermoelectrics: materials with high potential impact on energy conversion technology 2 Thermoelectrics are not in common use: low efficient and expensive 3 Thermolectrics are mainly present in niche markets (e.g. space technology) where reliability and simplicity are more critical issues than cost Minnich et al., Energy Environ. Sci. 2, 466 (2009) Key quantity: the thermoelectric figure-of-merit σ ZT = S2 T κ SixGe1-x bulk alloy The best thermoelectric material: MIN κ & MAX σ not provided by Nature! Metals have high electrical conductivity, but are excellent heat conductors as well Glasses have very low thermal conductivity, but are also poor charge conductors Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Search for new thermoelectrics with improved figure-of-merit 1 new chemically-complex materials with tailored physical REVIEW properties ARTICLE − CoSb3 − Doped CoSb3 − Ru0.5Pd0.5Sb3 − FeSb2Te − CeFe3CoSb12 SiGe CeFe3CoSb12 2 l (W m–1 K–1 ) 8 3 l (W m–1 K–1 ) UNICA 6 4 2 0 100 Hf0.75Zr0.25NiSn Bi2Te3 Sb 200 300 400 Temperature °C 500 Co Zn Yb 1 PbTe TAGS Ag9TlTe5 La3–xTe4 Yb14MnSb11 0 0 Zn4Sb3 200 400 Ba8Ga16Ge30 600 800 Temperature °C Figure 2 Complex crystal structures that yield low lattice thermal conductivity. a, Extremely low thermal conductivities are found in the recently identified complex material systems (such as Yb14MnSb11, ref. 45; CeFe3CoSb12, ref. 34; Ba8Ga16Ge30, ref. 79; and Zn4Sb3, ref. 80; Ag9TlTe5, ref. 40; and La3–xTe4, Caltech unpublished data) compared with most state-of-the-art thermoelectric alloys (Bi2Te3, Caltech unpublished data; PbTe, ref. 81; TAGS, ref. 69; SiGe, ref. 82 or the half-Heusler alloy Hf0.75Zr0.25NiSn, ref. 83). b, The high thermal conductivity of CoSb3 is lowered when the electrical conductivity is optimized by doping (doped CoSb3). The thermal conductivity is further lowered by alloying on the Co (Ru0.5Pd0.5Sb3) or Sb (FeSb2Te) sites or by filling the void spaces (CeFe3CoSb12) (ref. 34). c, The skutterudite structure is composed of tilted octahedra of CoSb3 creating large void spaces shown in blue. d, The room-temperature structure of Zn4Sb3 has a crystalline Sb sublattice (blue) and highly disordered Zn sublattice containing a variety of interstitial sites (in polyhedra) along with the primary sites (purple). e, The complexity of the Yb14MnSb11 unit cell is illustrated, with [Sb3]7– trimers, [MnSb4]9– tetrahedra, and isolated Sb anions. The Zintl formalism describes these units as covalently bound with electrons donated from the ionic Yb2+ sublattice (yellow). Snyder et al., Nature Materials 7, 105 (2008) 2 new nanostructured materials characterized by a large number of interfaces, efficiently working as phonon scatters superlattices Venkatasubramaniana et al., Nauture 413, 597 (2001) Wright discusses how alloying Bi Te with other isoelectronic cations zT adds enough carriers to substantially reduce thermal conductivity through electron–phonon (Fig. 2b). Further reductions and anions does not reduce the electrical conductivity but et lowers the Science quantum dot arrays Harman al., 297,interactions 597 (2002) thermal conductivity . Alloying the binary tellurides (Bi Te , Sb Te , can be obtained by alloying either on the transition metal or the PbTe and GeTe) continues to be an active area of research . Many antimony site. nanowires Hochbaum et al., Nature 451, 163 (2008) of the recent high-zT thermoelectric materials similarly achieve a Filling the large void spaces with rare-earth or other heavy atoms 2 28 34 3 2 3 29–32 2 3 reduced lattice thermal conductivity through disorder within the unit cell. This disorder is achieved through interstitial sites, partial further reduces the lattice thermal conductivity35. A clear correlation has been found with the size and vibrational motion of the filling rattling atomscost in additionof to the fabrication: disorder inherent in atom and the thermal conductivity leading to zT values as high as 1 Still a common feature occupancies, is theor high the alloying used in the state-of-the-art materials. For example, rare- (refs 8,13). Partial filling establishes a random alloy mixture of filling earth chalcogenides18 with the Th3P4 structure (for example La3–xTe4) atoms and vacancies enabling effective point-defect scattering as have a relatively low lattice thermal conductivity (Fig. 2a) presumably discussed previously. In addition, the large space for the filling atom due to the large number of random vacancies (x in La3–xTe4). As in skutterudites and clathrates can establish soft phonon modes and phonon scattering by alloying depends on the mass ratio of the alloy local or ‘rattling’ modes that lower lattice thermal conductivity. constituents, it can be expected that random vacancies are ideal Filling these voids with ions adds additional electrons that scattering sites. require compensating cations elsewhere in the structure for charge The potential to reduce thermal conductivity through disorder balance, creating an additional source of lattice disorder. For the case within the unit cell is particularly large in structures containing void of CoSb3, Fe2+ frequently is used to substitute Co3+. An additional spaces. One class of such materials are clathrates8, which contain benefit of this partial filling is that the free-carrier concentration large cages that are filled with rattling atoms. Likewise, skutterudites7 may be tuned by moving the composition slightly off the chargesuch as CoSb3, contain corner-sharing CoSb6 octahedra, which can balanced composition. Similar charge-balance arguments apply to be viewed as a distorted variant of the ReO3 structure. These tilted the clathrates, where filling requires replacing group 14 (Si, Ge) with octahedra create void spaces that may be filled with rattling atoms, as group 13 (Al, Ga) atoms. shown in Fig. 2c with a blue polyhedron33. Claudio Melis Atomistic Modeling of nanostructured materials For skutterudites containing elements with low electronegativity COMPLEX UNIT CELLS chemically-complex materials are expensive because the use rare elements nanostructured materials are typically grown by nanofabrication processes → unpractical for large-scale commercial use for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Search for new thermoelectrics with improved figure-of-merit New nanostructured materials characterized by a large number of interfaces, efficiently working as phonon scatters superlattices Venkatasubramaniana et al., Nature 413, 597 (2001) quantum dot arrays Harman et al., Science 297, 2229 (2002) nanowires Hochbaum et al., Nature 451, 163 (2008) SiGe superlattice STEM image of Si/SiGe nanowires Wu, Fan, Yang, Nano Lett. 2, 83 (2002) Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Semicondutor nanocomposites 1 Concept: nanocomposite grain size smaller that the phonon mean free path grain size larger that charge-carrier mean free path 2 Fabrication: bulk processes (ball milling, hot pressing) create thermally stable systems (thermoelectrics must operate for years at high temperature) low-cost processing 3 Materials: semiconductors abundant and low-cost benefit of well-established technology can tune electron conduction as needed Strategies for next-generation nc-based thermoelectrics TEM image of Si 0.8 Ge0.2 nanopowder 1 increasing charge-carrier mobility (reduce impact of grain boundaries on electron transport) 2 minimizing lattice thermal conductivity → basic motivation of the present work 3 reducing the electronic thermal conductivity Joshi et al. Nano Lett. 8, 4670 (2008) Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results An alternative approach: APPROACH-TO-EQUILIBRIUM MD (AEMD) Melis et al., EPJ-B 87, 96 (2014) 1 a periodic step-like initial temperature profile is assigned Formal solution of the heat equation in PBC provides T1 z T2 0 ∞ T(z; t) = Lz Lz /2 X 2 T1 + T2 + Bn sin(αn z)e−αn κ̄t 2 n=1 simulation cell 2 during the following transient thermal conduction we calculate T1,ave (t) = 1 Lz /2 Lz /2 Z T(z; t)dz and T2,ave (t) = 0 1 Lz /2 Z Lz T(z; t)dz Lz /2 from which the time-dependent average temperature difference is defined as ∆T(t) = T1,ave (t) − T2,ave (t) 3 it can be shown that ∆T(t) = ∞ X 2 Cn e−αn κ̄t with κ̄ = κ/ρCv n=1 Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Actual implementation of an AEMD simulation is really straightforward! 1 set-up a PBC step-like temperature profile → age the system by a NVE simulation 2 evaluate on-the-flight the time-dependent average temperature difference ∆T(t) P −α2n κ̄t → get κ̄ and so the thermal conductivity fit ∆T(t) = ∞ n=1 Cn e 300 500 t=0 ps t=20 ps t=50 ps t=100 ps 450 simulation analytical solution 250 400 200 350 ∆T (K) 3 Temperature [K] UNICA 300 150 100 250 50 200 150 0 0 20 40 60 80 100 120 140 160 0 z [nm] 50000 100000 150000 200000 250000 300000 350000 400000 time (fs) Sample with Lz =503.2 nm (in PBC) Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results A benchmark calculation: thermal conductivity in bulk c-Si at 600K interaction potential: EDIP Justo et al. PRB 58, 2539 (1998) simulation cell: Lx × Ly × Lz with Lz Lx = Ly initial temperature profile: T(z; t = 0) periodic step-like function with T1 − T2 = ∆T0 200 0.05 3x3x200 5x5x200 7x7x200 Lx,y = 3a0 Lx,y = 5a0 Lx,y = 7a0 150 ue 0.045 in g of Lz l va 100 1/K (W/m/K)-1 !T (K) as re c In 0.04 50 0.035 0.03 Lz = 200a0 0 0 10000 0.025 20000 30000 time (fs) 40000 50000 Extrapolated value of thermal conductivity 0.02 0 0.0002 0.0004 0.0006 0.0008 0.001 1/Lz (nm)-1 200 !T(0)=200 K !T(0)=100 K !T(0)=50 K ∆T0 = 200K ∆T0 = 100K ∆T0 = 50K 150 !T (K) UNICA Results this work: κ = 51 ± 5 WK−1 m−1 100 expt.: 50 κ = 64 WK−1 m−1 Glassbrenner et al., PR 134, A1058 (1964) for 50 K ≤ ∆T0 ≤ 200K the result is affected by ∼ 15% 0 0 50000 100000 time (fs) 150000 200000 Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results STEP 1: Generating templates an amorphous slab with size 5 × 50 × 50 in units of a0 is generated by quenching from the melt N grains are inserted at random as 1 2 3 4 5 N sites are selected at random in the yz plane a cylindrical void is generated at each site by removing atoms void radii are assigned at random so as to: (i) avoid overlap; (ii) starting above the capillarity threshold each void is filled by a randomly-rotated crystalline cylinder configurations, differing in crystallinity, saved on-the-flight: hereafter referred to as templates Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results STEP 1: Generating templates an amorphous slab with size 5 × 50 × 50 in units of a0 is generated by quenching from the melt N grains are inserted at random as 1 2 3 4 5 N sites are selected at random in the yz plane a cylindrical void is generated at each site by removing atoms void radii are assigned at random so as to: (i) avoid overlap; (ii) starting above the capillarity threshold each void is filled by a randomly-rotated crystalline cylinder configurations, differing in crystallinity, saved on-the-flight: hereafter referred to as templates Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results STEP 1: Generating templates an amorphous slab with size 5 × 50 × 50 in units of a0 is generated by quenching from the melt N grains are inserted at random as 1 2 3 4 5 6 N sites are selected at random in the yz plane a cylindrical void is generated at each site by removing atoms void radii are assigned at random so as to: (i) avoid overlap; (ii) starting above the capillarity threshold each void is filled by a randomly-rotated crystalline cylinder long annealing at the temperature of interest (1200K) configurations, differing in crystallinity, saved on-the-flight: hereafter referred to as templates Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results STEP 1: Generating templates an amorphous slab with size 5 × 50 × 50 in units of a0 is generated by quenching from the melt N grains are inserted at random as 1 2 3 4 5 N sites are selected at random in the yz plane a cylindrical void is generated at each site by removing atoms void radii are assigned at random so as to: (i) avoid overlap; (ii) starting above the capillarity threshold each void is filled by a randomly-rotated crystalline cylinder configurations, differing in crystallinity, saved on-the-flight: hereafter referred to as templates Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results STEP 1: Generating templates an amorphous slab with size 5 × 50 × 50 in units of a0 is generated by quenching from the melt N grains are inserted at random as 1 2 3 4 5 N sites are selected at random in the yz plane a cylindrical void is generated at each site by removing atoms void radii are assigned at random so as to: (i) avoid overlap; (ii) starting above the capillarity threshold each void is filled by a randomly-rotated crystalline cylinder configurations, differing in crystallinity, saved on-the-flight: hereafter referred to as templates Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results STEP 2: Generating nc samples a number M of templates are glued together so as to generate a sample as long as M × 50 a0 further long-time annealing typically: 1.5 - 2.0×106 time-step samples are aged until full recrystallization: hereafter referred to as nc samples This work: 2 ≤ M ≤ 7 → 30 nm < Lz < 200 nm number of atoms up to ∼ 7 × 105 Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results STEP 3: Generating Si1−x Gex nanocomposites lattice sites of nc samples are randomly decorated by Si and Ge atoms at the given stoichiometry x atomic positions are rescaled (through a self-affine transformation) according to the alloy lattice constant: Ge a0 (x) = x aSi 0 + (1 − x) a0 further relaxation provides samples to use for thermal transport investigations: hereafter referred to as Si1−x Gex nanocomposites This work: x = 0.2, 0.4, 0.8 This 3-step procedure is very computer-effective since: 1 STEP 1 is moderately computer-intensive but must be executed just once 2 STEP 2 is very computer-intensive but must be executed just once 3 STEP 3 is computationally light and must be repeated for any given stoichiometry x Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Bulk Si1−x Gex alloy 250 This work Ref.[13] Ref.[20] Ref.[21] Ref.[22] κ [WK-1m-1] 200 This work 150 1 Si1−x Gex modeled by Tersoff potential 2 196000 ≤ N ≤ 392000 atoms 3 272nm ≤ Lz ≤ 543nm Key issues 100 50 0 0 0.2 0.4 0.6 0.8 1 % Germanium Content 1 κ dramatically reduced 2 same reduction obtained for 0.2 ≤ x ≤ 0.8 3 minimum κ achieved by alloying 4 more than 50% of κ is due to phonons with λ ≥ 1µm Colors: expt. data Black: ab initio calcs. Garg et al. PRL 106, 045901 (2011) Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Bulk Si1−x Gex nanocomposite 1 κ vs. Ge-content for hdg i = 12.5 nm samples 6 Bulk alloy Nanocomposite κ [WK-1m-1] 5 !dg " = 2.5 nm 4 3 2 1 0.2 2 !dg " = 12.5 nm 0.3 0.4 0.5 0.6 0.7 % Germanium Content 0.8 κ vs. hdg i 4 Nanocomposite Bulk alloy 3.5 κ [WK-1m-1] 3 !dg " = 25.0 nm 2.5 2 1.5 1 0.5 0 0 Claudio Melis 5 10 15 20 25 Average grain size [nm] 30 Atomistic Modeling of nanostructured materials for energy and biomedical applic Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Thermal conductivity accumulation κ(L) κbulk 0.65 0.6 0.55 0.5 0.45 0.4 EXTRAPOLATED BULK VALUE 0.35 0 0.001 0.002 0.003 0.004 -1 1/Lz [nm ] 1 memo :κ ∼ vCV λm.f .p. Thermal conductivity accumulation [%] T.C.A. = 1/κ [W-1Km] UNICA 100 90 Bulk alloy Nanocomposite 80 70 60 50 40 30 20 10 0 0.0001 0.001 0.01 0.1 1 Phonon mean free path [µm] 10 at 300 K heat carried mostly by “phonons” with mean free paths ≤ 100 nm ∼ 60% of κ carried by short-propagating “phonons” with mean free path shorter than 50 nm 2 grain boundaries effectively reduce thermal conductivity by affecting the “phonon” mean free path through additional scattering Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic UNICA Example 1: Intro Ex. 1: MD simulations Ex 2: Elastic properties Example 2: Intro Example 2: AEMD Ex. 2: nc-SiGe models Ex. 2: Results Methods - Transient conduction MD simulations 1 robust theoretical foundation allow for investigating quite different boundary conditions (... even other than PBC) 2 ease of implementation no need to compute heat current no need to establish a steady-state situation 3 comparatively light computational effort only transient evolution is needed - no convergence 4 numerically very stable with respect to cross section choice of ∆T0 P −α2n κ̄t actual number of exponentials used in the fit ∆T(t) = ∞ n=1 Cn e Physics - Thermal conductivity in Si1−x Gex nanocomposites 1 present proof-of-concept simulations show that nc-SiGe has thermal conductivity below the alloy limit 2 thermal conductivity marginally depends on stoichiometry largely affected by granulometry (i.e. by hdgrain i) Claudio Melis Atomistic Modeling of nanostructured materials for energy and biomedical applic