# Molecular Dynamics - University of Washington Molecular Dynamics Simulations Joo Chul Yoon with Prof. Scott T. Dunham Electrical Engineering University of Washington Contents Introduction to MD Simulation Setup Integration Method Force Calculation and MD Potential

MD Simulations of Silicon Recrystallization Simulation Preparation SW Potential Tersoff Potential Introduction to Molecular Dynamics Calculate how a system of particles evolves in time Consider a set of atoms with positions /velocities and the potential energy function of the system Predict the next positions of particles over some short time interval by solving Newtonian mechanics

Basic MD Algorithm Set initial conditions ri (t0 ) and v i (t0 ) Get new forces Fi (ri ) Solve the equations of motion numerically over a short step t ri (t ) ri (t t ) v i (t ) v i (t t ) t t t Is t t max ? Calculate results and finish Simulation Setup

Simulation Cell Boundary Condition Constructing neighboring cells Initial atom velocities MD Time step Temperature Control Simulation Cell usually using orthogonal cells Open boundary for a molecule or nanocluster in vacuum not for a continuous medium Fixed boundary

fixed boundary atoms completely unphysical Periodic boundary conditions obtaining bulk properties Periodic boundary conditions An atom moving out of boundary comes back on the other side considered in force calculation L rcut

2 rcut Constructing neighboring cells 2 pair potential calculation O( N ) atoms move 0.2 A per time step not necessary to search all atoms Verlet neighbor list containing all neighbor atoms within rL

rcut updating every N L time steps where rL rcut N L vt 2 i rL skin

Constructing neighbor cells Linked cell method divide MD cell into smaller subcells : The length of subcell l L l rL n n n n is chosen so that L : the length of MD cell

going through 27 NN C atom pairs instead N ( N 1)! 3 where N C N / n rL reducing it to O( N ) 26 skin cells Simulation Setup Simulation Cell

Boundary Condition Constructing neighboring cells Initial atom velocities MD Time step Temperature Control Initial Velocities Maxwell-Boltzmann distribution The probability of finding a particle with speed m P( v x )

2k BT 1/ 2 1 exp mv 2x / k BT 2 Generate random initial atom velocities scaling T with equipartition theorem 3

1 k BT mv 2 2 2 MD Time Step Too long t : energy is not conserved r/t 1/20 of the nearest atom distance

In practice t 4 fs. MD is limited to <~100 ns Temperature Control Velocity Scaling Scale velocities to the target T Efficient, but limited by energy transfer Larger system takes longer to equilibrate Nose-Hoover thermostat Fictitious degree of freedom is added Produces canonical ensemble (NVT) Unwanted kinetic effects from T oscillation

Integration Method Finite difference method Numerical approximation of the integral over time Verlet Method Better long-tem energy conservation Not for forces depending on the velocities Predictor-Corrector Long-term energy drift (error is linear in time) Good local energy conservation (minimal fluctuation) Verlet Method From the initial ri (t ) , v i (t ) 1

a(r ) F (r (t )) m Obtain the positions and velocities at t t 1 r (t t ) r (t ) v(t )t a(r )t 2 2 1 a(t t ) a(r (t t )) m 1 v(t t / 2) v(t )t a(r )t 2 1

v(t t ) v(t t / 2) a(t t )t 2 Predictor-Corrector Method Predictor Step from the initial ri (t ) , v i (t ) 1 a(r ) F (r (t )) m predict ri (t t ) , v i (t t ) using a Taylor series a(t ) 2 r (t t ) r (t ) v(t )t t

2 v P (t t ) v(t ) a(t )t P a P (t t ) a(t ) r iii (t )t r iii : 3rd order derivatives Predictor-Corrector Method Corrector Step get corrected acceleration P F (

r (t t )) C a (r ) m using error in acceleration a(t t ) a C (t t ) a P (t t ) correct positions and velocities 2 t r (t t ) r P (t t ) C0 a(t t )

2 v(t t ) v P (t t ) C1ta(t t ) Cn : constants depending accuracy Force Calculation The force on an atom is determined by N u (rij ) j i rij

Fi U ( r ) rij U (r ) : potential function N : number of atoms in the system rij : vector distance between atoms i and j

MD Potential Classical Potential U U1 (ri ) U 2 (ri , r j ) U 3 (ri , r j , rk ) ... i i, j i , j ,k U1 : Single particle potential Ex) external electric field, zero if no external force

U 2 : Pair potential only depending on Fij r U (rij ) i U 3 : Three-body potential with an angular dependence Fi i (Vij V ji ) V jki j k j Using Classical Potential

Born-Oppenheimer Approximation Consider electron motion for fixed nuclei ( me M 0) Assume total wavefunction as (R i , r ) (R i ) (r , R i ) (R i ) : Nuclei wavefunction

(r , R i ) : Electron wavefunction parametrically depending on R i The equation of motion for nuclei is given by Pi 2 H N U (R i ) i 2M i (approximated to classical motion) MD Potential Models Empirical Potential

functional form for the potential fitting the parameters to experimental data Ex) Lennard-Jones, Morse, Born-Mayer Semi-empirical Potential calculate the electronic wavefunction for fixed atomic positions from QM Ex) EAM, Glue Model, Tersoff Ab-initio MD direct QM calculation of electronic structure Ex) Car-Parrinello using plane-wave psuedopotential Stillinger-Weber Potential works fine with crystalline and liquid silicon

U U 2 (ri , r j ) U 3 (ri , r j , rk ) i, j i , j ,k U 2 (rij ) f 2(rij / ) U 3 (ri , r j , rk ) f3 (ri / , r j / , rk / ) , : energy and length units Pair potential function

f 2(r ) A( Br p r q ) exp[(r a ) 1 ] , r a 0 , r a Stillinger-Weber Potential Three body potential function f3 (ri , r j , rk ) h(rij , rik , jik ) h(rji , rjk , ijk )

h(rki , rkj , ikj ) h(rij , rik , jik ) exp[ (rij a) 1 (rik a ) 1 ] 1 (cos jik ) 2 3 Stillinger-Weber Potential 1 2 (cos )

Limited by the cosine term jik 3 forces the ideal tetrahedral angle not for various equilibrium angles too low coordination in liquid silicon incorrect surface structures incorrect energy and structure for small clusters Bond-order potential for Si, Ge, C bond strength dependence on local environment Tersoff, Brenner

Tersoff Potential cluster-functional potential environment dependence without absolute minimum at the tetrahedral angle The more neighbors, the weaker bondings U U repulsive ( rij ) bijkU attractive (rij ) bijk : environment-dependent parameter weakening the pair interaction when coordination number increases Tersoff Potential

U ij f C (rij )[aij f R (rij ) bij f A (rij )] where repulsive part f R ( r ) Ae 1r attractive part f A ( r ) Be 2 r potential cutoff function f C (r )

1, r R D 1 1 (r R) sin , R D r R D 2 2 D

2 0, r R D Tersoff Potential bij (1 n ijn ) 1/ 2 n ij f C (rik ) g ( jik ) exp[33 (rij rik )3 ] k i , j

c2 c2 g ( ) 1 2 2 d d (h cos ) 2 aij (1 nijn ) 1/ 2 n ij f C (rik ) exp[33 (rij rik )3 ] k i , j Contents

Introduction to MD Simulation Setup Integration Method Force Calculation and MD Potential MD Simulations of Silicon Recrystallization Simulation Preparation SW Potential Tersoff Potential MD Simulation Setup Initial Setup

5 TC layer 1 static layer 4 x 4 x 13 cells MD Simulation Setup System Preparation Ion Implantation(1 keV) Cooling to 0K

Recrystallization 1200 K for 0.5 ns Recrystallization SW Potential 1200K Crystal Rate a/c interface displacement MD Simulation Setup

Initial Setup 6 TC layer 1 static layer 5 x 5 x 13 cells MD Simulation Setup System Preparation

Ion Implantation(1 keV) Cooled to 0K Recrystallization 1900 K for 0.85 ns Recrystallization Tersoff Potential 1900K Crystal Rate

a/c interface displacement Recrystallization Crystal Rate SW Potential 1200K Tersoff Potential 1900K Recrystallization a/c interface displacement SW Potential 1200K

Tersoff Potential 1900K MD Simulation Setup Initial Setup 6 TC layer 1 static layer 2 x 2 x 13 cells

Recrystallization 1800 K for 20 ns Tersoff Potential Melting temperature of Tersoff: about 2547K Potential energy per particle versus temperature: the system with a/c interface is heated by adding energy at a rate of 1000K/ns Tersoff Potential As in recrystallized Si :

o2 0.82 A2 / 100 ps in amorphized Si o A / 100 ps in crystalline Si 0.20 Tersoff Potential As in recrystallized Si : o2 A / 100 ps 0.82

in amorphized Si 2 o 0.20 A / 100 ps in crystalline Si Summary Review Molecular Dynamics MD simulation for recrystallization of Si with SW, Tersoff with As

## Recently Viewed Presentations

• VFTLP+ Re-reflected Energy Original Reflections Improved Reflections 5 ns long voltage pulse followed by reflected pulse at DUT VFTLP+ Re-reflected Energy Original Reflections Improved Reflections 5 ns long voltage pulse followed by reflected pulse at DUT * * * *
• Interface metaphors & analogies pp 58 - 63 pp 55 - 60 Interface metaphors Designed to be similar to a physical entity but also has own properties e.g. desktop metaphor, search engine Exploit user's familiar knowledge, helping them to understand...
• Spatial contiguity Temporal contiguity Severity-Area-Duration Highest severities Weibull percentiles Figure 6. Overview of methodology from initial input to envelope curves. 5 Envelope Curves Soil Moisture Runoff Sep 1976 to Nov 1976 2 Hydrologic Simulations and Validation Dec 1976 to Feb...
• Self-Portraits A Look into Artists Throughout History Self-Portraits Albrecht Durer Self-Portrait at Twenty Seven Oil on Wood Panel 20.5" x 16" Self-Portraits John Steuart Curry Self Portrait Drawing 25.5" x 19.5" Self-Portraits Jean-Baptiste Simeon Chardin Self Portrait in Spectacles Pastel...
• "Airtight" (it means what you intend) Auditable (it identifies an agreed-upon methodology) Enforceable (it specifies how damages are to be calculated - and paid - if the PBM breaches the guarantee) You Must Demand A Contract with Numerous Guarantees, &...
• needed a cleaner mapping from the logical units of computation to . each physical process. The importance of such clean mapping for . debug-ability is really crucial when . responding to pager alerts for a . failing topology, especially if...
• Assessment of Hops for Tennessee. David Hughes and Hannah Wright, UTIA Extension. Agricultural & Resource Economics. Outline. Growth in Craft Beer-Hops. Nature of plant, processing. Hops in the Southeast. Hops. Hops is a flower of Humulus lupulus .
• HOW CAN NEUROIMAGING HELP UNDERSTAND, DIAGNOSE, AND DEVELOP TREATMENTS FOR ALZHEIMER'S DISEASE? Part C - AD brain scans - anatomical NUCLEAR MEDICINE GRAND ROUNDS Stanford University J. Wesson Ashford, M.D., Ph.D. Clinical Professor (affiliated), Department of Psychiatry and Behavioral Sciences