Computational Physics Project •Predictor-Corrector method •Verlet Integration Guy Halioua 302748546 Technion – Israel Institute of Technology Preface – Molecular Dynamics Simulation method for exploring Dynamic systems Based on Newton’s laws of motion Solve problems with multiple bodies General Algorithm for MD Determine primary Positions and velocities of the system, and choose a small ∆t Calculate accelerations by: F V aF M Calculate location and velocity of every particle in the system at the next time step Update time to the next time step, and repeat for as much as you need. Integration Methods Solve a second order ODE: x f x, x, t Predictor Corrector - Multiple values method Verlet Integration – Numerical approximations to the derivatives Predictor - Corrector Predict a primitive guess for the values of x t t and x t t by the formulas: P x x t t x t t x t t k 1 2 f t 1 i t i i 1 P x t x t t x t t x t t k 1 2 ' f t 1 i t i 1 While: k 1 1 i q i 1 k 1 i 1 q 1 q 2 1 1 i i ' q 2 i 1 q i q 0,1,..., k 2 Predictor - Corrector Calculate the accelerations for time t+∆t based on the prediction made earlier. x t t f x t t , x t t , t Predictor - Corrector Correct the primitive guess for the values of x t t and x t t made O t k 1 earlier by the formulas: C x x t t x t t x t t k 1 2 f t 2 i t i i 1 C x t x t t x t t x t t k 1 2 ' f t 2 i t i 1 k 1 While: 2 i q i 1 k 1 2 i i 1 q i i 1 q 1 q 2 1 i ' q 2 q 0,1,..., k 2 Predictor - Corrector Predictor Corrector coefficients for secondorder equations Taken from D.C Rapaport’s Book: “The Art of Molecular Dynamics Simulation” Verlet Integration Taylor Expand of x t t and x t t t 2 x t t x t v t t a t O t 3 2 t 2 x t t x t v t t a t O t 3 2 x t t 2 x t x t t t 2 a t O t 4 Verlet Integration While a t can be found from Newton’s equations of motion: a t F x t M V x t M And the velocity is found by the mean value theorem: x t t x t t v t O t 2 2t Verlet Integration Accuracy analysis: Velocity O t 2 Overall accuracy O t 2 4 O t Coordinate Looking for a better approximation Velocity Verlet Integration Simply a Taylor Expand of x t t and v t t t 2 x t t x t v t t a t O t 3 2 1 v t t v t t a t a t t 2 while a t t is taken from the motion equation using x t t MD Simulation 1D Row of Linear Oscillators Defining the problem: row of masses, divided by linear oscillators (k,l) 1 2 3 4 n2 n 1 n Determine the Dynamics of the system MD Simulation 1D Row of Linear Oscillators Defining the potential function: Elastic potential k 2 ui xi xi 1 l 2 n K U xi xi 1 l i 1 2 Force U Fi k xi 1 xi 1 2 xi xi MD Simulation 1D Row of Linear Oscillators Choosing parameters: m 1 kg Mass a 0.5 m Lattice’s const’ k 1 N m l 0.5 a 0.25 m Oscillator’s const’ n 30 No. of particles t 5 104 [sec] Size of time step T 40000 No. of time steps Oscillator’s free length MD Simulation 1D Row of Linear Oscillators Choosing initial conditions: Coordinates: Lattice organization around the origin Velocities: randomly picked, range [-0.3,0.3] using Matlab’s random number generator MD Simulation 1D Row of Linear Oscillators Programming Two programs in C for computing the positions, velocities and energies at all the time steps. Input is a file with the initial velocities MD Simulation 1D Row of Linear Oscillators Expected results: Solution of 1 oscillator – sine (cosine) wave Expected solution for multiple oscillators, multiple sine (cosine) waves. MD Simulation 1D Row of Linear Oscillators Results - Predictor Corrector MD Simulation 1D Row of Linear Oscillators Results – Verlet Integration MD Simulation 1D Row of Linear Oscillators Results – animation: Verlet Integration Predictor – Corrector MD Simulation 1D Row of Linear Oscillators Results – Differences: MD Simulation 1D Row of Linear Oscillators Results – Differences: Max Difference ≈ 1.2e-4 [m] Typical system size- a=0.5 m % max difference: 0.024% Identical Results MD Simulation 1D Row of Linear Oscillators Energy analysis – Predictor Corrector MD Simulation 1D Row of Linear Oscillators Energy analysis – Verlet Integration MD Simulation 1D Row of Linear Oscillators Conclusions Energy and momentum conservation Identical results in 2 different methods Compatible with theory Logical Results