Slide 1

Slide 1 text

1 Introduction Molecular Dynamics Method: Theory and Implementation Department of Applied Physics and Physico-Informatics, Keio University Hiroshi Watanabe

Slide 2

Slide 2 text

2 Molecular Dynamics Simulation Integrate equations of motion numerically • Calculate Force • Update momenta • Update position

Slide 3

Slide 3 text

3 q = 0.0 p = 1.0 h = 0.01 t = 0.0 for i in range(1000): q += p * h p -= q * h t += h print(t, q, p) Python code of MD

Slide 4

Slide 4 text

4 Particle methods ARE NOT MD. Governing Eq. MD Equations of Motion SPH Navier-Stokes Eq. CFD Navier-Stokes Eq. Specification Eulerian Lagrangian Lagrangian

Slide 5

Slide 5 text

5 Range Force Ab Initio (DFT) Classical Short-Range (LJ, WCA, etc) Long-Range (Coulomb, Gravity etc)

Slide 6

Slide 6 text

6 Boiling from Wikipedia from Wikipedia Cavitation Bubbles are important. • Bubbles affect efficiency of power plants. • Bubbles cause noise and damage on propeller. Understand behavior of bubbles and control them.

Slide 7

Slide 7 text

7 Spontaneous Phase Transition Stable for creating/annihilation of Phase Boundary Size of Simulation is highly limited. ~ um^3

Slide 8

Slide 8 text

8 Bubbles Gas/Liquid Surface Atoms Micro ~nm Macro ~cm Huge-scale Molecular Dynamics Simulation Bubbles are difficult. • Multi-scale and multi-physics problem • Moving, creation, and annihilation of gas-liquid surface Investigate the behavior of bubbles from the atomic scale

Slide 9

Slide 9 text

9 F = ma Very simple Easy to understand

Slide 10

Slide 10 text

10 F = ma Very simple Easy to understand Really?

Slide 11

Slide 11 text

11 What is temperature? What is pressure? Observables Time Evolution What is time evolution? What is time integration?

Slide 12

Slide 12 text

12 Extended Hamiltonian Methods Andersen’s method to control pressure Nose’s method to control temperature How can we control extensive variables? What is physical meaning of an extended system? We have the least action principle for original dynamics. What we have for the extended dynamics?

Slide 13

Slide 13 text

13 Variable We KNOW variables a priori. Used without definition. Observable We DEFINE observables. Defined by using variables.

Slide 14

Slide 14 text

14 Heat Equation = ∆ Temperature is Variable Navier-Stokes Equation Ԧ = − + ∆ Ԧ Pressure is Variable Equation of Motion ሶ = − , ሶ = Temperature is Observable Pressure is Observable

Slide 15

Slide 15 text

15 Equation of Motion ሶ = − , ሶ = Temperature is Observable Pressure is Observable Momenta are Variables. Coordinates are Variables. Observables should be defined by Variables.

Slide 16

Slide 16 text

16 We usually adopt the following temperature. = 2 3 ෍ 2 2 We usually adopt the following pressure. = 2 3 + 1 3 ෍ < ∙ …But Why?

Slide 17

Slide 17 text

17 We usually use Symplectic Integrators. Symplectic Integrator is better than other scheme. … But Why? Can we apply SI for extended systems?

Slide 18

Slide 18 text

18 We want to control pressure and/or temperature. We can control pressure by Andersen’s method. We can control temperature by Nose’s method. … But Why? Why can we control pressure and/or temperature? What happen in the extended system? Extended System

Slide 19

Slide 19 text

19 We consider the fundamental aspects of MD. We will consider • What is the time evolution? • What is the time integration? • What is pressure? How can we control? • What is temperature? How can we control?

Slide 20

Slide 20 text

20 We consider the implementation of MD. We will consider • Memory optimization • SIMD vectorization • Programming Design

Slide 21

Slide 21 text

21 0. Introduction 1. Classical Mechanics 2. Pressure 3. Temperature 4. Numerical Integration 5. Nose-Hoover method 6. Langevin Thermostat 7. Integration scheme for non-Hamiltonian systems 8. Generalized Liouville Theorem 9. Implementation and Optimization 10. Programming Design Theory Implementation

Slide 22

Slide 22 text

22 https://kaityo256.github.io/md2019/ "md2019" Lecture Note (written in Japanese)