Introduction to Geometry Optimization. Computational Chemistry lab 2009

Similar documents
Project: Vibration of Diatomic Molecules

Exploring the energy landscape

Example questions for Molecular modelling (Level 4) Dr. Adrian Mulholland

Density Functional Theory

2. The Schrödinger equation for one-particle problems. 5. Atoms and the periodic table of chemical elements

Reactivity and Organocatalysis. (Adalgisa Sinicropi and Massimo Olivucci)

Simulations with MM Force Fields. Monte Carlo (MC) and Molecular Dynamics (MD) Video II.vi

Molecular Simulation II. Classical Mechanical Treatment

Unconstrained Multivariate Optimization

MD Thermodynamics. Lecture 12 3/26/18. Harvard SEAS AP 275 Atomistic Modeling of Materials Boris Kozinsky

From Electrons to Materials Properties

Molecular Bonding. Molecular Schrödinger equation. r - nuclei s - electrons. M j = mass of j th nucleus m 0 = mass of electron

Normal Mode Analysis. Jun Wan and Willy Wriggers School of Health Information Sciences & Institute of Molecular Medicine University of Texas Houston

Figure 1: Transition State, Saddle Point, Reaction Pathway

Chemistry 334 Part 2: Computational Quantum Chemistry

Optimization Methods via Simulation

Today. Introduction to optimization Definition and motivation 1-dimensional methods. Multi-dimensional methods. General strategies, value-only methods

Chemistry 3502/4502. Final Exam Part I. May 14, 2005

Lecture 35 Minimization and maximization of functions. Powell s method in multidimensions Conjugate gradient method. Annealing methods.

8 Numerical methods for unconstrained problems

Same idea for polyatomics, keep track of identical atom e.g. NH 3 consider only valence electrons F(2s,2p) H(1s)

An Introduction to Quantum Chemistry and Potential Energy Surfaces. Benjamin G. Levine

Statistical Geometry Processing Winter Semester 2011/2012

Chemistry 3502/4502. Final Exam Part I. May 14, 2005

Quantum Theory of Matter

Computer simulation can be thought as a virtual experiment. There are

Lecture 8 Optimization

Methods of Computer Simulation. Molecular Dynamics and Monte Carlo

( ) 2 75( ) 3

Making Electronic Structure Theory work

Numerical Optimization: Basic Concepts and Algorithms

Physics 403. Segev BenZvi. Numerical Methods, Maximum Likelihood, and Least Squares. Department of Physics and Astronomy University of Rochester

( ) + ( ) + ( ) = 0.00

Molecular energy levels

Chapter 10 Conjugate Direction Methods

Gradient Descent. Dr. Xiaowei Huang

Chemistry 324, Spring 2002 Geometry Optimization

TECHNIQUES TO LOCATE A TRANSITION STATE

Geometry optimization

Conjugate gradient algorithm for training neural networks

OPER 627: Nonlinear Optimization Lecture 14: Mid-term Review

Unconstrained optimization

1 Kernel methods & optimization

DFT and beyond: Hands-on Tutorial Workshop Tutorial 1: Basics of Electronic Structure Theory

Topic 8c Multi Variable Optimization

Molecular Mechanics. I. Quantum mechanical treatment of molecular systems

5.1 Classical Harmonic Oscillator

(One Dimension) Problem: for a function f(x), find x 0 such that f(x 0 ) = 0. f(x)

, b = 0. (2) 1 2 The eigenvectors of A corresponding to the eigenvalues λ 1 = 1, λ 2 = 3 are

The Conjugate Gradient Method

L20: MLPs, RBFs and SPR Bayes discriminants and MLPs The role of MLP hidden units Bayes discriminants and RBFs Comparison between MLPs and RBFs

A6523 Signal Modeling, Statistical Inference and Data Mining in Astrophysics Spring

Numerical Optimization Professor Horst Cerjak, Horst Bischof, Thomas Pock Mat Vis-Gra SS09

Comparison of methods for finding saddle points without knowledge of the final states

Transition states and reaction paths

Minimization of Static! Cost Functions!

Multivariate Newton Minimanization

Geometry Optimisation

Phonons and lattice dynamics

Vasil Khalidov & Miles Hansard. C.M. Bishop s PRML: Chapter 5; Neural Networks

Rotations and vibrations of polyatomic molecules

22 Path Optimisation Methods

Molecular Mechanics, Dynamics & Docking

Intro/Review of Quantum

Chapter 6. Nonlinear Equations. 6.1 The Problem of Nonlinear Root-finding. 6.2 Rate of Convergence

Harmonic Oscillator Eigenvalues and Eigenfunctions

Intro/Review of Quantum

LECTURE # - NEURAL COMPUTATION, Feb 04, Linear Regression. x 1 θ 1 output... θ M x M. Assumes a functional form

Example - Newton-Raphson Method

Optimization. Next: Curve Fitting Up: Numerical Analysis for Chemical Previous: Linear Algebraic and Equations. Subsections

The Potential Energy Surface (PES) Preamble to the Basic Force Field Chem 4021/8021 Video II.i

EFFICIENT SAMPLING OF SADDLE POINTS WITH THE MINIMUM-MODE FOLLOWING METHOD

A Primer on Multidimensional Optimization

Steepest descent algorithm. Conjugate gradient training algorithm. Steepest descent algorithm. Remember previous examples

Report due date. Please note: report has to be handed in by Monday, May 16 noon.

= F ( x; µ) (1) where x is a 2-dimensional vector, µ is a parameter, and F :

Programming, numerics and optimization

A dimer method for finding saddle points on high dimensional potential surfaces using only first derivatives

Methods that avoid calculating the Hessian. Nonlinear Optimization; Steepest Descent, Quasi-Newton. Steepest Descent

ECE 595, Section 10 Numerical Simulations Lecture 7: Optimization and Eigenvalues. Prof. Peter Bermel January 23, 2013

Improving the Convergence of Back-Propogation Learning with Second Order Methods

Chemistry 431. NC State University. Lecture 17. Vibrational Spectroscopy

Spectroscopy: Tinoco Chapter 10 (but vibration, Ch.9)

Functions of Several Variables

Dispersion relation for transverse waves in a linear chain of particles

ECE580 Exam 2 November 01, Name: Score: / (20 points) You are given a two data sets

Outline. Escaping the Double Cone: Describing the Seam Space with Gateway Modes. Columbus Workshop Argonne National Laboratory 2005

What is Classical Molecular Dynamics?

MOLECULAR STRUCTURE. The general molecular Schrödinger equation, apart from electron spin effects, is. nn ee en

Density Functional Theory: from theory to Applications

(2 pts) a. What is the time-dependent Schrödinger Equation for a one-dimensional particle in the potential, V (x)?

Handout 10. Applications to Solids

Molecular Dynamics. Park City June 2005 Tully

/2Mα 2 α + V n (R)] χ (R) = E υ χ υ (R)

ECE580 Exam 1 October 4, Please do not write on the back of the exam pages. Extra paper is available from the instructor.

Introduction to gradient descent

Mathematical optimization

G : Statistical Mechanics

Tools for QM studies of large systems

Neural Network Training

Transcription:

Introduction to Geometry Optimization Computational Chemistry lab 9

Determination of the molecule configuration H H Diatomic molecule determine the interatomic distance H O H Triatomic molecule determine distances and the angle H N H H Four-atomic molecule determine angles and distances Optimal configurations correspond to minima of the energy as a function of 3N- 6 parameters for an N-atomic molecule (3N atomic coordinates-3 c.o.m coordinates - 3 angles)

Solution of the Schrödinger equation in the Born-Oppenheimer approimation: H tot = (T n + V n ) + T e + V ne + V e = (H n ) + H e lectronic S: H e Ψ (r,r)= e (R) Ψ (r,r) Ab-initio or semi-empirical solution e (R) electronic energy R nuclear coordinates V n (R) niclei-nuclei interaction (R)=V n (R)+ e (R) = potential energy surface (PS) Molecular mechanics: PS molecular force field

Potential nergy Surface (PS) A hypersurface in 3N-6 -dimensional hyperspace, where N is the number of atoms. Minima equilibrium structures of one or several molecules Saddle points - transition structures connecting equilibrium structures. Global Minimum - that point that is the lowest value in the PS

One configuration is nown for H One configuration is nown for H O The planar and pyramidal configurations are nown for NH 3 The number of local minima typically goes eponentially with the number of variables (degrees of freedom). Combinatorial plosion Problem Possible Conformations (3 n ) for linear alanes CH 3 (CH ) n+ CH 3 n = 3 n = 9 n = 5 43 n = 59,49 n = 5 4,348,97

Global minimum - Mariana Trench in the Pacific Ocean

Optimization/Goals Find the Local Minimum Structure Find the Global Minimum Structure Find the Transition State Structure We would lie to Find the Reaction Pathway Connecting Two Minima and Passing through the Transition State Structure Geometry optimization is the name for the procedure that attempts to find the configuration of minimum energy of the molecule.

One-dimensional optimization f 3 maimum f f ()= stationary point f ()< maimum minimum 5 f ()> minimum f' - f'' -5 -.5..5..5..5

Multidimensional Optimization Force: F=- (Potential energy) - gradient The coordinates can be Cartesian (,y,z..), or internal, such as bondlength and angle displacements) f= - Stationary Point (minimum, maimum, or saddle point)

Hessian matri A matri of second-order derivatives of the energy with respect to atomic coordinates (e.g., Cartesian or internal coordinates) Sometimes called force matri matri size of (3N-6)(3N-6) ( H ij ( f ) f i j = Quadratic approimation Approimate the comple energy landscape by harmonic potentials around a stationary point (,... () () () () n [ (,... n ) = ] n n () () = + () (),... n) (,... n ) Hij( )( i i )( j j ) i= j= )

Hessian matri diagonalization (eigenproblem) igenvalues igenvectors All ε > ( ω real) minimum ω ε l ( ) j ε = mω - vibrational frequencies n ( ) ( l j give normal coordinates q = li () () (,... ) = (,... + ε q n n ) n = All ε < ( ω imaginary) maimum i= = ) n i= ( i H ij l ( ) i () i ε <> ( ω imaginary/real) saddle point )

Steepest Descent Steepest descent direction g=- Ε g - 3N-dimensional vector = + λ g / g i+ i i i i Advantage: fast minimization far from the minimum Disadvantage: slow descent along narrow valleys slow convergence near minima In HYPRCHM if ( i + ) < ( i) then i+ =.λi else λi+ = History gradient information is not ept λ. 5λ i

Conjugate Gradient Methods Search direction is chosen using history gradient information Initial direction the steepest descent direction, h = g =- Ε = + ξh i+ i i ξ is defined by a one-dimensional minimization along the search direction. hi+ = gi+ + γ i+ hi g i Fletcher-Reeves method: γ + i+ = gi ( i+ gi ) g Pola-Ribiere method: γ i+ = g g i+ Pola-Ribiere may be superior for nonquadratic functions. i steepest conjugate Gradient history is equivalent to implicit use of the Hessian matri.

plicit use of Hessian matri. Quadratic approimation : ( ) = q ε - normal coordinates, and l Newton-Raphson Methods - ( = eigenvalues and eigenvectors of = n q l ), = g n ( F q = n = F l ε q the Hessian matri at, ), One-step optimization of quadratic functions q = F / ε For arbitrary functions n i+ = i + l ( i) F ( i ) / ε ( i = Steeprst descent direction for ε ( i ) > ) Newton- Raphson conjugate Finds the closest stationary point (either minimum, maimum, or saddle point).

Numerical calculation (3N) calculations of energy Diagonalization - (3N) 3 operations Memory - (3N) tremely epensive for large molecules. Obtaining, storing, and diagonalaizing the Hessian Bloc Diagonal Newton-Raphson z z y z z y y y z y z z y z z y y y z y Calculation - 9N Diagonalization - 7N Memory 9N

ε Gaussian The Hessian matri is calculated and processed on the first step only It is updated using the computed energies and gradients Rational function optimization (RFO) step i+ = and l Berny algorithm i + n = l ( i F ( i ) ) ε ( ) λ λ<min(ε ) (λ< for ε <) - step always toward to a minimum. A linear search between the latest point and the best previous point. i eigenvalues and eigenvectors of the Hessian matri at RFO Newton-Raphson

Convergence criteria Gradient or force F=-g=- Ε Predicted displacement i+ - i Maimum force: ma(f)<threshold Root-mean-square (RMS) force: F <threshold Maimum displacement: ma( i+ - i ) <threshold RMS displacement: i+ - i <threshold

One configuration is nown for H One configuration is nown for H O The planar and pyramidal configurations are nown for NH 3 The number of local minima typically goes eponentially with the number of variables (degrees of freedom). Combinatorial plosion Problem Possible Conformations (3 n ) for linear alanes CH 3 (CH ) n+ CH 3 n = 3 n = 9 n = 5 43 n = 59,49 n = 5 4,348,97

Global minimization "building up" the structure: combining a large molecule from preoptimized fragments (protein from preoptimized aminoacids) conformational sampling: tae various starting points for local minimization (simulation of nature) None of these are guaranteed to find the global minimum!

Global Minimum- conformational sampling Molecular Dynamics Monte Carlo Simulated Annealing Tabu search Genetic Algorithms Ant Colony Optimizations Diffusion Methods Distance Geometry Methods

Heating above barriers System overcomes barriers at sufficiently high temperature T quilibrium (Boltzmann) distribution P(R)~ep(-(R)/( B T )) Higher probability to find the system near the global minimum

Molecular Dynamics Solve equations of motion d T v j = j( R) + ( ) dt m j τ T R j atomic coordinates d v j = R j atomic velocities dt 3 T = v j temperature B j m j T bath temperature τ bath relaation time The trajectory fills available phase space Monte Carlo. One atom is randomly "iced". The energy change = new - old is calculated 3. If < the new configuration is accepted 4. If > the new configuration is accepted with probability ep(- /( B T )) 5. Net atom is randomly "iced" Both methods lead to equilibrium averaged distribution in the limits of infinite time or number of steps

Simulated annealing If a slow cooling is applied to a liquid, this liquid freezes naturally to a state of minimum energy. bath temperature heating equilibration cooling ither Molecular-Dynamics or Monte-Carlo simulation can be used. Is guaranteed to find the global minimum if the equilibration is infinitely long and the cooling is infinitely slow.