Scribd Logo
Upload

Welcome to Scribd!

  • 54 views

    Molecular Dynamics Simulations: Che 520 Fall 2009

    Uploaded by

    Fabio Okamoto
    Molecular dynamics (MD) simulations involve numerically solving Newton's equations of motion to simulate particle interactions over time. Key steps include initializing particle positions and velocities, calculating interparticle forces, and integrating equations of motion using algorithms like Verlet. MD is similar to Monte Carlo methods in modeling small systems with periodic boundaries but differs in directly simulating dynamic particle evolution over time. Time step sizes are limited to around 1-2 femtoseconds to resolve fastest vibrations and minimize long-term energy drift from numerical errors.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd

    Molecular Dynamics Simulations: Che 520 Fall 2009

    Uploaded by

    Fabio Okamoto
    54 views8 pages
    Molecular dynamics (MD) simulations involve numerically solving Newton's equations of motion to simulate particle interactions over time. Key steps include initializing particle positions and velocities, calculating interparticle forces, and integrating equations of motion using algorithms like Verlet. MD is similar to Monte Carlo methods in modeling small systems with periodic boundaries but differs in directly simulating dynamic particle evolution over time. Time step sizes are limited to around 1-2 femtoseconds to resolve fastest vibrations and minimize long-term energy drift from numerical errors.

    Original Title

    MD1

    Copyright

    © Attribution Non-Commercial (BY-NC)

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Molecular dynamics (MD) simulations involve numerically solving Newton's equations of motion to simulate particle interactions over time. Key steps include initializing particle positions and velocities, calculating interparticle forces, and integrating equations of motion using algorithms like Verlet. MD is similar to Monte Carlo methods in modeling small systems with periodic boundaries but differs in directly simulating dynamic particle evolution over time. Time step sizes are limited to around 1-2 femtoseconds to resolve fastest vibrations and minimize long-term energy drift from numerical errors.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    54 views8 pages

    Molecular Dynamics Simulations: Che 520 Fall 2009

    Uploaded by

    Fabio Okamoto
    Molecular dynamics (MD) simulations involve numerically solving Newton's equations of motion to simulate particle interactions over time. Key steps include initializing particle positions and velocities, calculating interparticle forces, and integrating equations of motion using algorithms like Verlet. MD is similar to Monte Carlo methods in modeling small systems with periodic boundaries but differs in directly simulating dynamic particle evolution over time. Time step sizes are limited to around 1-2 femtoseconds to resolve fastest vibrations and minimize long-term energy drift from numerical errors.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 8

    ChE 520 Fall 2009

    Molecular Dynamics Simulations

    http://www.ks.uiuc.edu/Gallery/Movies/ This movie illustrates the conduction of KCL through a silica nanopore. The nanopore has 1 nm radius, KCL is in solution and under 1 V bias
    2

    Molecular Dynamics (MD): main idea


    Setup system in initial configuration Solve Newtons equations of motion:

    d 2ri
    2 dt

    Fi mi

    i = 1,2,..., N

    Measure properties after system has reached equilibrium, e.g. Energy, pressure, pair correlation functions
    3

    Similarities and differences with Monte Carlo (MC)


    Both MD and MC: use small systems, periodic boundary conditions, sample from relevant statistical mechanical ensembles subject to time scale/ergodicity constraints Differences: Natural ensemble for MD is microcanonical (const.-E), MC easier at const. T Actual dynamic evolution of system studied in MD, phase space sampling for MC Direct calculation of transport coeff. in MD MC more flexible, unphysical moves can be used to accelerate sampling
    4

    Simple MD code: 1. Initialization


    Particle positions initialized at random or on a lattice Velocities need to be assigned: At equilibrium, mean-square velocity along any coordinate: kT 2 < v2 >=< v 2 >=< v z >= x y m Equilibrium velocity distribution is MaxwellBoltzmann:
    mv 2 exp 2
    5

    P( v ) = (2mkT )3/2

    Initialization (cont.)
    No need to start from Maxwell-Boltzmann, velocities are quickly randomized
    subroutine init ! initialization of MD program sumv(1:3) = 0. ! sum of velocities along each coordinate sumv2 = 0. ! sum of squares of velocities do i=1,Npart do j = 1,3 xyz(j,i) = ran(seed)*L ! L is the box length v(j,i)= ran(seed)-0.5 ! random velocity sumv(j) = sumv(j) + v(j,i) sumv2 = sumv2 + v(j,i)*v(j,i) enddo enddo sumv(1:3) = sumv(1:3)/Npart fs = sqrt(3*(Npart-1)*T/sumv2) ! because COM does not move ! set COM motion to zero and mean kinetic T to desired value v(1:3,1:Npart) = fs*(v(1:3,1:Npart) - sumv(1:3)) rm(1:3,1:Npart) = xyz(1:3,1:Npart) - v(1:3,1:Npart)*dt return end!
    6

    Initialization (cont.)
    As equilibration proceeds, temperature is going to drift away from its desired value. The mean kinetic temperature of the system at any time is:
    Tkin (t ) = i = 1 3k( N 1)
    N

    miui2(t )

    The factor N 1 appears because the center of mass (COM) is fixed in space. The velocity rescaling shown on slide 5 can be applied at any time; however, it is not a proper way to achieve constant-temperature conditions and it does not conserve energy.
    7

    Simple MD code: 2. Moving the particles


    We first need to evaluate the forces acting on each particle. In general, the force components are:
    f x (r ) = x 2 + y 2 + z 2 U(r ) U(r ) r U(r ) x U(r ) = = = x x r x r r r

    For the LJ potential (with ==1), this becomes:


    f x (r ) = 48 x 1 1 r 2 r 12 2r 6

    The force calculation code segment is executed millions and millions of times, so be very careful how you write it!
    8

    Force calculation code


    ! Determine force f and energy en f = 0. ; en = 0. ! Array and scalar set to 0 do i=1,Npart-1 do j = i+1,Npart dr(1:3) = xyz(1:3,i) xyz(1:3,j) ! 1=x, 2=y, 3=z dr(1:3) = dr(1:3) nint(dr(1:3)*aL) ! aL is 1/L r2 = dr(1)*dr(1) + dr(2)*dr(2) + dr(3)*dr(3) r2i = 1./r2 ; r6i = r2i*r2i*r2i ff(1:3) = r2i*r6i*(r6i-0.5) do idim=1,3 f(idim,i) = f(idim,i) + ff(idim)*dr(idim) f(idim,j) = f(idim,j) - ff(idim)*dr(idim) enddo en = en + r6i*(r6i-1.) enddo enddo f = 48.*f ; en = 4.*en

    No cutoff or shift is applied to the code above (will lead to instability and loss of energy conservation)
    9

    Moving the particles (cont.)


    Once the forces have been calculated, the equations of motions can now be integrated forward in time. There are many possible numerical integration schemes. Molecular dynamics integrators must satisfy the following conditions: calculation of forces is expensive and can only be done once per time step long-term energy conservation is important time reversibility and conservation of volume in phase space are required (symplectic integrators) Some of the simplest methods are also the best.
    10

    Moving the particles (cont.): The Verlet algorithm


    f (t ) 2 t 3 r(t + t ) = r(t ) + v(t )t + t + r + O( t 4 ) 2m 3 f (t ) 2 t 3 r(t t ) = r(t ) v(t )t + t r + O( t 4 ) 2m 3 summing these two expressions: f (t ) 2 r(t + t ) + r(t t ) = 2r(t ) + t + O( t 4 ) m subtracting : r(t + t ) r(t t ) v(t ) = + O( t 2 ) 2 t
    Positions are more accurately known than velocities

    11

    Moving the particles (cont.): Velocity Verlet


    f (t ) 2 t 2m f (t ) + f (t + t ) v(t + t ) = v(t ) + t 2m r(t + t ) = r(t ) + v(t )t +
    This is equivalent to the original Verlet for the positions, but has better short-term energy conservation because of more accurate velocity calculations

    12

    Time steps and energy drift


    For any time step, numerical errors that accumulate result in long-term energy drift an exponential increase in energy for very large number of integration steps. For Verlet integration, the practical limit for the maximum time step is

    2 t =

    Where is the fastest vibrational time scale in the system (see link). For typical soft matter systems with bending / torsional potentials, this is approximately 1 - 2 fs, depending on whether explicit-hydrogens and fixed bond lengths are used. For the LJ system, the reduced time is: [ For Ar, the unit of time is ~2 ps ]

    t t* = /m
    13

    Typically, t* = 0.005 for stability [= 10 fs for Ar]

    Liapunov Instability

    ( ) r(t ) = f ( r (0), p (0) + ,t )


    r(t ) = f r N (0), pN (0),t
    N N

    r r = r(t )

    For short times, r(t ) For long times, r(t ) exp( t ) : Liapunov exponent (positive)
    14

    15

    You might also like

    pFad - Phonifier reborn

    Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

    Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


    Alternative Proxies:

    Alternative Proxy

    pFad Proxy

    pFad v3 Proxy

    pFad v4 Proxy