CPMD Tutorial Atosim/RFCT 2009/10

Transcription

1 These exercices were inspired by the CPMD Tutorial of Axel Kohlmeyer axel.kohlmeyer/cpmd-tutor/index.html and by other tutorials. Here is a summary of what we will do: I Introduction to CPMD First, copy the input files to your directory: 1 mkdir CPMD cd CPMD cp -r ~userchimie/atosim_cpmd/inputs/*. source./cpmd.sh Each time you open a new window, you will have to type: cd CPMD source./cpmd.sh 1. Let s have a look at the CPMD code using H 2 (a) cd H2 (b) We will first optimize the wavefunction using H2_optwfn.inp, given below (you don t have to type it, it is already in the directory!): &INFO isolated hydrogen molecule. single point calculation. &CPMD OPTIMIZE WAVEFUNCTION CONVERGENCE ORBITALS 1.0d-7 &SYSTEM SYMMETRY 1 ANGSTROM CELL CUTOFF 70.0 &DFT FUNCTIONAL LDA 1 Each line is one instruction, so the following lines should be entered one after each other. 1

2 &ATOMS *H_MT_LDA LMAX=S Type cpmd.x H2_optwfn.inp > H2_optwfn.out and wait for a few seconds. (c) Look at the output file using your favorite editor vi or emacs. We will discuss it togeter. 2. We will now look at some molecular dynamics (MD)! (a) First, we will optimize the geometry of the H 2 molecule using: cpmd.x H2_optgeom.inp > H2_optgeom.out & While the calculation is running, have a look at the input file! (b) You can use molden to have a look at the file GEOMETRY.xyz to see the optimized geometry for H 2. (c) Copy the RESTART.1 file to save it: cp RESTART.1 RESTART.geomopt (d) Let s go for dynamics: cpmd.x H2_MD1.inp > H2_MD1.out & (e) While the calculation is running, have a look at the input file. &INFO isolated hydrogen molecule. simple molecular dynamics deltat=4au &CPMD MOLECULAR DYNAMICS CP RESTART WAVEFUNCTION COORDINATES LATEST TRAJECTORY XYZ MAXSTEP 100 TIMESTEP 4.0 &SYSTEM SYMMETRY 1 ANGSTROM CELL CUTOFF 2

3 70.0 &DFT FUNCTIONAL LDA &ATOMS *H_MT_LDA LMAX=S New keywords: TRAJECTORY XYZ, TIMESTEP. (f) Use molden to have a look at the trajectory file TRAJEC.xyz. To see the evolution, clisk on the Movie button. (g) What s happening? (h) Delete files that are not needed anymore: rm ENERGIES TRAJECTORY TRAJEC.xyz (i) Copy the restart file: cp RESTART.geomopt RESTART.1 (j) Use now cpmd.x H2_MD2.inp > H2_MD2.out & (k) Use molden to have a look at the trajectory file TRAJEC.xyz. Are things ok now? (l) Let s have a look at the file ENERGIES using gnuplot. The columns are in the folowing order: NFI A counter EKINC Kinetic energy of the electrons TEMPP The instantaneous temperature EKS Kohn-Sham energy (electronic energy) ECLASSIC Equivalent of the total energy in a classical MD run EHAM Total energy, should be conserved. ECLASSIC=EHAM-EKINC DIS Mean squared displacement of the atoms from the initial coordinates TCPU CPU time needed for this step. In gnuplot, type: 2 plot "ENERGIES" u 1:2 t "EKINC" w l plot "ENERGIES" u 1:3 t "TEMPERATURE" w l plot "ENERGIES" u 1:4 t "EKS" w l plot "ENERGIES" u 1:5 t "ECLASSIC" w l plot "ENERGIES" u 1:6 t "EHAM:conserved?" w l What are your conclusions? 3. We will now test the adiabaticity hypothesis by launching different calculations with different TIMESTEP and EMASS values. We can change the timestep between 2 and 20 a.u. and the fictitious electronic mass between 200 and The last caracter is an l not a one. 3

4 (a) Copy the restart file: cp RESTART.geomopt RESTART.1 before each calculations to always start for the same wavefunction and geometry. These calculations will be split between all of you. You will edit the H2_MD3.inp file to change the values of these two parameters. Somebody should also launch a BO simulation using H2_BOMD.inp. Be careful that the total time of the simulation should be the same. That is TimeStep*MaxStep should always be equal to 5000 atomic time units. Conclusions? II More realistic systems In this part, we wil consider two systems that are a bit more realistic: water, amonia and bulk silicon. 1 Water There are two directories, depending on how the periodic boundaries are treated: for H2O, there is no decoupling, ie a molecule can see its images, in H2O_isolated, we use a decoupling scheme. Choose one of them and compare to your neighbour. Repeat what we have done before: 1. Geometry optimization: H2O_optgeom.inp. Save the restart file: cp RESTART.1 RESTART.geomopt. Here, we ask to print some geometric parameters. Compare between the isolated calculation and the periodic one. 2. Short MD: H2O_MD1.inp with TimeStep=4, EMass=400 or H2O_MD2.inp with TimeStep=16, EMass=1200. We can also have a look at the IR spectrum, but this will needs a longer MD run. Launch H2O_MD_IR.inp. Do not change the RESTART file, or erase some files. As this will take some time, launch it as a background task while you are looking at the next calculations. During the MD, the evolution of the DIPOLE moment has been printed in the file DIPOLE. We can thus look at the IR spectrum because it is the Fourier Transform of the autocorrelation function of the dipole moment. For this, we will use the program fourier.x. Follow the instruction. 2 Amonia Here, we will focus on the flip-flop of the NH 3 molecule at 300K. For this, we will launch a NVT simulation using a Nosé-Hoover thermostat. Change to the directory NH3: cd../nh3 1. Optimize the geometry: NH3_optgeom.inp 2. Equilibrate the system at 300K: NH3_equil_300.inp or at 700K NH3_equil_700.inp. Have a look at the input file while the simulation is running! 3. Launch the simulation using a thermostat at the right temperature (you might have to change it in the input file): NH3_MD.inp Use molden to look at the molecule along the simulation with the TRAJEC.xyz file. Is the molecule freely flipping at 300K? at 700K? 4

5 3 Bulk silicon We will use this solid system to have a look at the radial distribution function. Change to the directory Si8: cd../si8 1. Optimize the wavefunction: Si8_optwfn.inp 2. Equilibrate the system at 1500K: Si8_equil.inp 3. Launch the simulation using a thermostat: Si8_equil.inp Use molden to look at the molecule along the simulation with the TRAJEC.xyz file and to compute the radial distribution function. III Long simulations 1 Free energy profile We finish this set of exercices by the calculation of the free energy profile for the rotation of one methyl group of ethane around the single CC bond. All files are in FreeEnergy. Many ways can be used to compute this: Brute force: launch a very very very very... very long simulation, and may be you will see the angle moving out of the initial minimum. I practice it does not work, even though we know that this rotation is almost free at room temperature. The sample input file is called Ethane_BruteForce.inp. You have to modify it to: 1. Optimize the wave function 2. Launch a very long dynamic Thermodynamic integration: we will launch many constrained simulation in which the reaction coordinate will be the torsion angle, θ, between the two methyl groups: H 1 theta H 6 H 5 C 7 C 8 H 3 H 2 H 4 A good way to do so is to launch simulation for angles equal to 0, 20, 40, 60, 80, 100 and 120. For each simulation, we will then compute the free energy derivative, and then integrate it. A sample file is given in Ethane_TI_60.inp. You can use molden to generate different geometries. You can use the Ethane.xyz file if needed. In molden, geometry can be modified using the Zmat editor menu. For each simulation, you will need to: 1. Create a geometry with the given value of the dihedral angle, 2. optimize the wave function, 3. launch a long dynamic (at least steps), 5

6 4. compute the free energy derivative using the CONSTRAINT file. Umbrella sampling. We will proceed in two steps. First, using CPMD, we will use the torsion angle as a constraint to compute the potential energy profile for the rotation of the methyl group. This will give us an idea of the barrier to overcome, and thus some hints about the biaising potential. See the thermodynamic integration procedure to compute the energy here. For each value of the dihedral angle, you will need to: 1. Create a geometry with the given value of the dihedral angle, 2. Optimize the wave function, 3. Optimize the geometry (with the dihedral constrained). Add this potential using CPMD restraints, and launch a long simulation. A sample input file is Ethane_Umbrella_60.inp. An easy way is to use only one restraint centered on θ =60. How does it differ from the Brute force option? Metadynamic. This is a bit too long for us. Try it later if you have some time... 6