Advanced methods of molecular dynamics Monte Carlo methods

Slides:



Advertisements
Similar presentations
PRAGMA – 9 V.S.S.Sastry School of Physics University of Hyderabad 22 nd October, 2005.
Advertisements

Monte Carlo Simulation Wednesday, 9/11/2002 Stochastic simulations consider particle interactions. Ensemble sampling Markov Chain Metropolis Sampling.
Review Of Statistical Mechanics
Molecular Dynamics at Constant Temperature and Pressure Section 6.7 in M.M.
Lecture 13: Conformational Sampling: MC and MD Dr. Ronald M. Levy Contributions from Mike Andrec and Daniel Weinstock Statistical Thermodynamics.
Monte Carlo Methods and Statistical Physics
Introduction to Statistical Thermodynamics (Recall)
1 Lecture 6 Ideal gas in microcanonical ensemble. Entropy. Sackur-Tetrode formula. De Broglie wavelength. Chemical potential. Ideal gas in canonical ensemble.
1 CE 530 Molecular Simulation Lecture 8 Markov Processes David A. Kofke Department of Chemical Engineering SUNY Buffalo
The Calculation of Enthalpy and Entropy Differences??? (Housekeeping Details for the Calculation of Free Energy Differences) first edition: p
Monte Carlo Simulation Methods - ideal gas. Calculating properties by integration.
MSEG 803 Equilibria in Material Systems 8: Statistical Ensembles Prof. Juejun (JJ) Hu
Computational statistics 2009 Random walk. Computational statistics 2009 Random walk with absorbing barrier.
Computer Simulations, Nucleation Rate Predictions and Scaling Barbara Hale Physics Department and Cloud and Aerosol Sciences Laboratory, University of.
Machine Learning CUNY Graduate Center Lecture 7b: Sampling.
Statistical Mechanics
Introduction to Monte Carlo Methods D.J.C. Mackay.
1 CE 530 Molecular Simulation Lecture 7 David A. Kofke Department of Chemical Engineering SUNY Buffalo
Simulation of Random Walk How do we investigate this numerically? Choose the step length to be a=1 Use a computer to generate random numbers r i uniformly.
Monte Carlo Methods: Basics
Introduction to (Statistical) Thermodynamics
Monte-Carlo simulations of the structure of complex liquids with various interaction potentials Alja ž Godec Advisers: prof. dr. Janko Jamnik and doc.
1 Physical Chemistry III Molecular Simulations Piti Treesukol Chemistry Department Faculty of Liberal Arts and Science Kasetsart University :
Free energies and phase transitions. Condition for phase coexistence in a one-component system:
Computational Solid State Physics 計算物性学特論 第8回 8. Many-body effect II: Quantum Monte Carlo method.
1 Lesson 3: Choosing from distributions Theory: LLN and Central Limit Theorem Theory: LLN and Central Limit Theorem Choosing from distributions Choosing.
Machine Learning Lecture 23: Statistical Estimation with Sampling Iain Murray’s MLSS lecture on videolectures.net:
1 CE 530 Molecular Simulation Lecture 18 Free-energy calculations David A. Kofke Department of Chemical Engineering SUNY Buffalo
Basic Monte Carlo (chapter 3) Algorithm Detailed Balance Other points.
The Local Free Energy Landscape - a Tool to Understand Multiparticle Effects University of Leipzig, Institut of Theoretical Physics Group Molecular Dynamics.
Module 1: Statistical Issues in Micro simulation Paul Sousa.
1 CE 530 Molecular Simulation Lecture 6 David A. Kofke Department of Chemical Engineering SUNY Buffalo
Monte Carlo Methods in Statistical Mechanics Aziz Abdellahi CEDER group Materials Basics Lecture : 08/18/
Nathan Baker BME 540 The Monte Carlo method Nathan Baker BME 540.
Simulated Annealing.
For a new configuration of the same volume V and number of molecules N, displace a randomly selected atom to a point chosen with uniform probability inside.
1 Statistical Mechanics and Multi- Scale Simulation Methods ChBE Prof. C. Heath Turner Lecture 14 Some materials adapted from Prof. Keith E. Gubbins:
8. Selected Applications. Applications of Monte Carlo Method Structural and thermodynamic properties of matter [gas, liquid, solid, polymers, (bio)-macro-
Monte Carlo Methods So far we have discussed Monte Carlo methods based on a uniform distribution of random numbers on the interval [0,1] p(x) = 1 0  x.
Suppressing Random Walks in Markov Chain Monte Carlo Using Ordered Overrelaxation Radford M. Neal 발표자 : 장 정 호.
Study of Pentacene clustering MAE 715 Project Report By: Krishna Iyengar.
Advanced methods of molecular dynamics 1.Monte Carlo methods 2.Free energy calculations 3.Ab initio molecular dynamics 4.Quantum molecular dynamics 5.Trajectory.
Exploring the connection between sampling problems in Bayesian inference and statistical mechanics Andrew Pohorille NASA-Ames Research Center.
Lecture 2 Molecular dynamics simulates a system by numerically following the path of all particles in phase space as a function of time the time T must.
Molecular Modelling - Lecture 2 Techniques for Conformational Sampling Uses CHARMM force field Written in C++
ChE 452 Lecture 17 Review Of Statistical Mechanics 1.
Path Integral Quantum Monte Carlo Consider a harmonic oscillator potential a classical particle moves back and forth periodically in such a potential x(t)=
7. Metropolis Algorithm. Markov Chain and Monte Carlo Markov chain theory describes a particularly simple type of stochastic processes. Given a transition.
Monte Carlo method: Basic ideas. deterministic vs. stochastic In deterministic models, the output of the model is fully determined by the parameter values.
An Introduction to Monte Carlo Methods in Statistical Physics Kristen A. Fichthorn The Pennsylvania State University University Park, PA
General considerations Monte Carlo methods (I). Averages X = (x 1, x 2, …, x N ) – vector of variables P – probability density function.
Monte Carlo in different ensembles Chapter 5
Generalized van der Waals Partition Function
Monte Carlo methods (II) Simulating different ensembles
Javier Junquera Importance sampling Monte Carlo. Cambridge University Press, Cambridge, 2002 ISBN Bibliography.
MSc in High Performance Computing Computational Chemistry Module Parallel Molecular Dynamics (i) Bill Smith CCLRC Daresbury Laboratory
1 Statistical Mechanics and Multi- Scale Simulation Methods ChBE Prof. C. Heath Turner Lecture 19 Some materials adapted from Prof. Keith E. Gubbins:
Introduction to Sampling Methods Qi Zhao Oct.27,2004.
Statistical Mechanics and Multi-Scale Simulation Methods ChBE
Basic Monte Carlo (chapter 3) Algorithm Detailed Balance Other points non-Boltzmann sampling.
Chapter 6: Basic Methods & Results of Statistical Mechanics
Computational Physics (Lecture 11) PHY4061. Variation quantum Monte Carlo the approximate solution of the Hamiltonian Time Independent many-body Schrodinger’s.
Computational Physics (Lecture 10) PHY4370. Simulation Details To simulate Ising models First step is to choose a lattice. For example, we can us SC,
Systematic errors of MC simulations Equilibrium error averages taken before the system has reached equilibrium  Monitor the variables you are interested.
The Monte Carlo Method/ Markov Chains/ Metropolitan Algorithm from sec in “Adaptive Cooperative Systems” -summarized by Jinsan Yang.
Computational Physics (Lecture 10)
Overview of Molecular Dynamics Simulation Theory
Common Types of Simulations
CE 530 Molecular Simulation
Presentation transcript:

Advanced methods of molecular dynamics Monte Carlo methods Free energy calculations Ab initio molecular dynamics Quantum molecular dynamics Trajectory analysis

1. Monte Carlo methods Direct MC: hit & miss method Importance sampling: The Metropolis method Isobaric MC Grand canonical MC Kinetic MC

Direct MC Normal integration methods (e.g., Simpson) impractical in many dimensions. Instead, Monte Carlo: Hit & miss method for estimating multidimensional integrals F =  f(x) dx. No inherent konwledge of f(x). Good when f(x) positively (or negatively) definite. Bad for oscillatory functions.  = 4 Nhit/Ntotal

Importance Sampling Random numbers chosen from a specific distribution (x) such that the function is evaluated in regions which make important contributions. Generating a Markov chain of states (functional values) f1, f2, f3, … which has a limiting distribution (x). In a Markov chain fn depends only on fn-1. fn linked to fn-1 by a transition probability pn-1,n Microscopic reversibility: fn pn,n-1 = fn-1 pn-1,n A. A. Markov (1856 - 1922)

Metropolis method From state with energy En-1 to state with energy En by randomly displacing a particle (or several particles, or all of them): If En < En-1 … accept If En > En-1 … generate a random number R, 0 < R < 1, if R < exp(-(En-En-1)/kT) …accept if R > exp(-(En-En-1)/kT) …reject (1915 - 1999) Ideal acceptance ratio ~50%: too small – too high rejection rate, no move; too large - too small steps, little move. Generates canonical ensemble with limiting distribution: exp(-E/kT)

Advantages/Disadvantages of MC + Simple; no need to evaluate forces, + Directly samples the (canonical) statistical ensamble; no need to invoke the ergodic theorem, Does not explicitely contain the time variable; principally impossible to evaluate time-dependent (equilibrium) properties such as correlation functions, - For complex potentials Monte Carlo sampling can often be less efficient than that of molecular dynamics.

Isobaric Monte Carlo Generates canonical ensemble with NpT is the usual experimental ensemble: Additional factor in the partition function Zp = 0 dV VN exp (-pV/kT) Modified Metropolis method: From state with energy En-1 to state with energy En by randomly displacing particles and changing the volume (or lnV). Changing volume means displacing all particles & changing long range corrections (Ewald). Generates canonical ensemble with limiting distribution: exp(-(E+pV)/kT+NlnV)

Grand Canonical Monte Carlo Fixed temperature T, volume V, and chemical potential μ, i.e., the free energy of inserting a particle. Additional factor in the partition function: Zμ = Σ0 (N!)-1 VN/Λ3 x exp(-Nμ/kT), … Λ: thermal wavelength Modified Metropolis method: From state with energy En-1 to state with energy En by randomly displacing particles and changing the number of particles by +/-1. Generates canonical ensemble with limiting distribution: exp(-(E-Nμ)/kT-lnN!-3NlnΛ+NlnV)!

Grand Canonical Monte Carlo II Implementations Simple-minded method method: Randomly switching particles from “existing“ to “ghost“ by changing ocupancy numbers (1 or 0). Then applying Metropolis method (ghost atom moves always accepted). More sophisticated algorithms: Different types of moves: (i) a particle is displaced, (ii) a particle is destroyed (no record kept), and (iii) a particle is created at a random position. Micorscopic reversibility by making the creation and destruction probabilities equal. Problems with high rejection rates (unfavorable overlaps when particle is created).

Grand Canonical Monte Carlo III Problems: In dense systems (fluids) it is hard to create a new particle without drastically increasing energy -> large rejection rate (special algorithms looking for cavites). Practical implementation – Widom insertion method: μ = -kT ln(QN/QN+1) μ = μideal gas + μexcess μexcess = -kT ln dsN+1 <exp(-(E(sN+1)-E(sN))/kT)>N - conventional NVT Monte Carlo with N particles, - frequent random insertions of an extra particle, - evaluation of exp(-(E(sN+1)-E(sN))/kT) & averaging

Grand Canonical Monte Carlo IV Movie

Kinetic Monte Carlo Allows to simulate time evolution. However, not at the molecular level but by introducing reaction rates (which have to be known from elsewhere, e.g., from transition state theory). At each step, system can jump from state A into one of the ending states Bi. survival probability: psurvival(t) = exp (-ktot t), ktot = ΣkABi integrated probability of escape between 0 and t: 1 – psurvival(t) Repeated many times – Markovian process, i.e., system looses memory before doing the next step. Most often used for surface diffusion or growth.

Kinetic Monte Carlo Procedure A stochastic algorithm propagating the system A -> B -> C… System is in state A, For each path using known escape probability pABi we generate a random transition time tBi We choose a path with shortest transition time tBmin We proceed to the next step. Advantages: detailed balance preserved, long (second) times accessible. Problems: system can visit states which were not intuitively expected and for which rate constant is not given, small barriers question valididty of the Markov chain and shorten the accesible time scale.