Chimie Théorique » Opt'n Path

\documentclass[a4paper,11pt]{article}

\usepackage{a4wide}

%\usepackage{times}

\usepackage{graphicx}

\usepackage[body={16cm,24.5cm}]{geometry}

\usepackage[latin1]{inputenc}

\usepackage[T1]{fontenc}

\usepackage{ae,aecompl}

%\input{m-pictex.tex}

%\usepackage{m-ch-en}

\usepackage{amsmath,amssymb}

%\usepackage{slashbox}

%\usepackage{psfrag}

%\usepackage{multirow}

%\usepackage{amscd}

%\usepackage{empheq}

%\usepackage{yhmath}

%\usepackage{array}

\usepackage{fancyhdr}

%\usepackage{braket}

\usepackage{marvosym}

%\usepackage{xr}

\DeclareGraphicsExtensions{.eps,.ps}

\graphicspath{{.}{../}}

\renewcommand{\arraystretch}{1.2}

\setcounter{tocdepth}{2}

\makeatletter

\@addtoreset{section}{part}

\makeatother

\lhead[]{}

\rhead[]{}

\cfoot{}

\title{OpenPath mini-help}

\author{P. Fleurat-Lessard, P. Dayal \\

Laboratoire de Chimie de l'ENS Lyon, 46 all?e d'Italie, F-69364 Lyon

Cedex 7}

\date{Feb. 2011}

\def\Path{\texttt{OpenPath}}

\begin{document}

\maketitle

\section{introduction}

\Path{} is a program that can optimize a reaction path between two

  structures. The algorithm to optimize the path is close to the string

  method. The originality of this program lies in the coordinate set it

 can use to generate and optimize the path.

This program is an independant program that calls standard electronic

  structure codes to get the energies and forces it needs to optimize

  the reaction path. For now, it is coupled to Gaussian, MOPAC2009,

  Vasp and Turbomole.

\section{Installation}

We suppose here that you will install \Path{} into a directory called

\texttt{OpenPath}. First, create the directory:

\begin{verbatim}

 mkdir OpenPath

 cd OpenPath

 mv ../OpenPath.tgz .

\end{verbatim}

Uncompress the archive:

\begin{verbatim}

 gunzip OpenPath.tgz

 tar -xvf OpenPath.tar

\end{verbatim}

You should now have this \texttt{Mini\_help.pdf} file and 4 directories (doc, src, utils, examples).

\section{Compilation}

Go to the directory  in which you have uncompressed the files.

Change to the \texttt{src} directory.

Edit the \texttt{Makefile} to change the \texttt{Machine} description

according to the compiler you want to use. Main choices are:

gfortran, g95, ifort, pgf, xlf and pathscale for now. You might also

have to check that the locations of

the libraries are ok.

Type \texttt{make}. You should now have a file called

\texttt{Path.exe} in this directory, as well as two utilities called

\texttt{xyz2scan} and \texttt{xyz2path} located in the \texttt{utils}

directories. You should copy all these executables to your

\texttt{~/bin} directory (or any place from which they can be executed).

\section{Use}

\subsection{Path calculation}

 To call \Path{}, you can type:

\verb=Path.exe Input_file Output_file=

The input file \texttt{Input\_file} is based on a namelist and looks like:

\begin{verbatim}

 &path

 nat=3, ! Number of atoms

 ngeomi=3, ! Number of initial geometries

 ngeomf=12, !Number of geometries along the path

 OptReac=.T., ! Do you want to optimize the reactants ?

 OptProd=.T., ! Optimize the products

 coord='zmat', ! We use Z-matrix coordinates

 maxcyc=31, ! Max number of iterations

 IReparam=2,! re-distribution of points along the path every 2 iterations

 ISpline=50, ! Start using spline interpolation at iteration 50

 Hinv=.T. , ! Use inverse of the Hessian internally (default: T)

 MW=T, ! Works in Mass Weighted coordiante (default T)

 PathName='Path_HCN_zmat_test', ! Name of the file used for path outputs

 prog='gaussian',! we use G03 to get energy and gradients

 SMax=0.1 ! Displacement cannot exceed 0.1 atomic units (or mass weighted at. unit)

 /

  3

 Energy :      0.04937364

 H     0.0000     0.0000     0.0340

 C     0.0000     0.0000     1.1030

 N     0.0000     0.0000     2.2631

  3

 Energy :      0.04937364

 H     0.0000     1.1000     1.1030

 C     0.0000     0.0000     1.1030

 N     0.0000     0.0000     2.2631

3

 CNH

H 0.000000    0.000000    3.3

C 0.000000    0.000000    1.1

N 0.000000    0.000000    2.26

%chk=Test

#P  AM1 FORCE

 HCN est bien

0,1

H 0.000000    0.000000    0.000000

C 0.000000    0.000000    1.000

N 0.000000    0.000000    3.00

\end{verbatim}

   Compulsory variables are:

\begin{description}

\item          NGeomi: Number of geometries defining the Initial path

\item          NGeomf: Number of geometries defining the Final path

\item         Nat   : Number of atoms

\end{description}

          Other options

\begin{description}

\item  Input: string that indicates the type of the input geometries.

          Accepted values are: Cart (or Xmol or Xyz) or Vasp

\item    Prog: string that indicates the program that will be used for energy and gradient calculations.

               Accepted values are: Gaussian, Vasp or Ext. \\

                In case of a Gaussian calculations, input must be set to Cart.

                One example of a gaussian input should be added at the end of the

                input file.See example file \texttt{Test\_HCN\_zmat\_g03.path}. \\

                In the case of a VASP calculation, if input is set to Cart, then

              the preamble of a VASP calculation should be added at the end of

              the input file. See example file \texttt{Test\_VASP\_cart.path}.

              In the case of a VASP calculation, one should also give value of the

              \texttt{RunMode} variable .

\item RunMode: This indicates wether \Path{} should use VASP

              routine to calculate the energy and gradient of the whole path

              or not. User has two options. One is to calculate the energy and

              gradient of each point sequentially. This is usefull when

              running on one (or two) processors. In this case,

              \texttt{RunMode} should be put to \texttt{SERIAL}.When running

              in parallel with 8 or more processors, one can use VASP to

              calculate simultaneously the energies and gradients for all

              points, as in a normal NEB calculation. In this case,

              \texttt{RunMode} must be set to \texttt{PARA}.

\item ProgExe: Name (with full path) of the executable to be used to get

  energies and gradients. For example, if VASP is used in parallel, one might

  have something like: \\

\verb!ProgExe='/usr/local/mpich/bin/mpirun -machinefile machine -np 8 ~/bin/VASP_46'!.

Another option that I use, is to put \verb!ProgExe='./run_vasp'! and I put every option needed to run VASP

into the \texttt{run\_vasp} file.

\item   PathName: Prefix used to save the path. Default is Path

\item   Poscar: string that will be used as the prefix for the different

           POSCAR files in a VASP calculations. Usefull only if PathOnly=.TRUE.,

           not used for internal calculations.

\item   IGeomRef: Index of the geometry used to construct the internal coordinates.

         Valid only for Coord=Zmat, Hybrid or Mixed

\item   Fact: REAL used to define if two atoms are linked.

         If $d(A,B) \leq fact*(rcov(A)+rcov(B))$, then A and B are considered Linked.

\item   debugFile: Name of the file that indicates which subroutine should print debug info.

\item   Coord: System of coordinates to use. Possible choices are:

          - CART (or Xyz): works in cartesian

          - Zmat: works in internal coordinates (Zmat)

          - Mixed: frozen atoms, as well as atoms defined by the

          'cart' array(see below) are describe in CARTESIAN, whereas the

          others are described in Zmat

          - Baker: use of Baker coordinates, also called delocalized internal coordinates

          - Hybrid: geometries are described in zmat, but the gradient are used in cartesian

\item   Step\_method: method to compute the step for optimizing a geometry; choices are:

          - RFO: Rational function optimization

          - GDIIS: Geometry optimization using direct inversion in the iterative supspace

\item    HUpdate: method to update the hessian. By default, it is Murtagh-Sargent

         Except for geometry optimization where it is BFGS.

\item   MaxCyc: maximum number of iterations for the path optimization

\item   Smax: Maximum length of a step during path optimization

\item   SThresh: Step Threshold to consider that the path is stationary

\item   GThresh: Gradient Threshold to consider that the path is stationary, only orthogonal part is taken

\item   FTan: We moving the path, this gives the proportion of the displacement tangent to the path

         that is kept. FTan=1. corresponds to the full displacement,

         whereas FTan=0. gives a displacement orthogonal to the path.

\item   IReparam: The path is not reparameterised at each iteration. This gives the frequency of reparameterization.

\item   ISpline: By default, a linear interpolation is used to generate the path.

            This option indicates the first step where spline interpolation is used.

\end{description}

          Arrays:

\begin{description}

\item   Rcov: Array containing the covalent radii of the first 80 elements.

         You can modify it using, \verb!rcov(6)=0.8!.

\item   Mass: Array containing the atomic mass of the first 80 elements.

\item   AtTypes: Name of the different atoms used in a VASP calculations.

          If not given, Path will read the POTCAR file.

\end{description}

          Flags:

\begin{description}

\item        MW:  Flag. True if one wants to work in Mass Weighted coordinates. Default=.TRUE.

\item       Renum: Flag. True if one wants to reoder the atoms in the

  initial order. default is .TRUE. most of the time.

\item       OptProd: True if one wants to optimize the geometry of the products.

\item       OptReac: True if one wants to optimize the geometry of the reactants.

\item       CalcEProd: if TRUE the product energy will be

  computed. Default is FALSE. Not Compatible with RunMode=Para".

\item       CalcEReac: if TRUE the reactants energy will be

  computed. Default is FALSE. Not Compatible with RunMode=Para".

\item       PathOnly:TRUE if one wants to generate the initial path, and stops.

\item       Hinv: if True, then Hessian inversed is used.

\item       IniHup: if True, then Hessian inverse is extrapolated

  using the initial path calculations.

\item       HupNeighbour: if True, then Hessian inverse is

  extrapolated using the neighbouring points of the path.

\item       FFrozen: True if one wants to freeze the positions of some atoms.

                If True, a \verb!&frozenlist! namelist containing the

                list of frozen atoms must be given.

                 If VASP is used, and frozen is not given

        here, the program will use the F flags of the input geometry

\item       FCart:  True if one wants to describe some atoms using cartesian coordinates.

               *** Only used in 'mixed' calculations. ***

             If True, a \verb!&cartlist! namelist containing the list of cart atoms must be given.

             By default, only frozen atoms are described in cartesian coordinates.

\item       Autocart: True if you want to let the program choosing the cartesian atoms.

\item       VMD: TRUE if you want to use VMD to look at the Path. Used only for VASP for now

\end{description}

More to come...      have a look at the files provided in the

\texttt{examples} directory. In particular, you will find there three

directories containing easy-to-use examples. That is, each directory contains

all the files you need to launch the application. For now, you have working

examples for:

\begin{enumerate}

\item \texttt{Gaussian}: Examples  using sequential call

  to Gaussian to calculate the energies and forces along the path.

\item \texttt{VASP}: Examples  using

  to VASP to calculate the energies and forces along the path.

As VASP can perform NEB calculations, Path can use it in two ways:

Serial or Parallel.

\begin{itemize}

\item[Serial]

In the Serial mode, Path uses Vasp to compute the energy and forces of

each image separately (as it does for Gaussian for example).

Therefore, the INCAR file should not contain any references to the

number of images (no IMAGES command!).

\item[Parallel]

In the Parallel mode, Path uses VASP to compute at once the energies

and forces for all the images (as VASP does for a NEB calculation).

Therefore, the INCAR file MUST contain the IMAGES command.

More, the directories 00, 01, ... should exist.

\item[Home or Scratch ?]

On our local cluster, it is adviced to perform the calculations on the

/scratch directory that is local rather than on the /home that is

mounted  via NFS and is thus slow.

However, when doing so, we sometimes had troubles with VASP, and

discovered that all files should be copied on the /scratch of all

machines.

Therefore, the scripts are a bit tricky.

In case your computer center is similar to ours, we provide 2 SGE

scripts:

\begin{itemize}

\item[run\_Path\_SGE\_home] performs the VASP calculations in the /home

directory

\item[run\_Path\_SGE\_scratch] creates the directories, copies the

  files and performs the VASP calculations in the local /scratch

  directories.

\end{itemize}

\end{itemize}

\item \texttt{Mopac}: Examples  using sequential call

  to MOPAC2009 to calculate the energies and forces along the path.

\item \texttt{Test}: Examples  using the analytical HCN potential

  energy surface to calculate the energies and forces along the path.

As this is fast, we also provide other Analysis tools.

See README files in the Cart and Zmat subdirectories.

\end{enumerate}

\subsection{Path analysis}

In order to analyse the path evolution, we provide some utilities, in

the \texttt{utils} directory. \texttt{xyz2scan} and \texttt{xyz2path}

have to be compiled. For this, go to either the \texttt{src} or the

\texttt{utils} directory and type \texttt{make utils}.

Here is a brief description of these utilities.

\begin{itemize}

\item \texttt{AnaPath} and/or \texttt{AnaPathref} \\

These scripts analyse a calculated path. They use \texttt{xyz2path} to

convert the cartesian coordinates saved by \Path{} into a data file

(called PathName.datl) that be plotted using the gnuplot files that are

also created. They all plot the path energy  but in different ways:

\begin{itemize}

\item[PathName\_l.gplot] plots the energy of the first iteration (the

  initial path) together with the energy of the following iterations,

  but one at a time.

\item[PathName\_l2.gplot]  plots the energy of the path for all

  iterations, with respect to the

curvilinear distance along the path.

\item[PathName\_l3.gplot]  plots the energy of the path for all

  iterations, with respect to the index of the images.

\end{itemize}

\item \texttt{xyz2path} and \texttt{xyz2scan} convert a bunch of

  cartesian coordinates into

  a data file. They are very similar: the only difference is the way

  they print their results. For \texttt{xyz2scan} the analyses are

  indiced using the geometry number whereas \texttt{xyz2path} computes

  the mass weighted distance between two geometries and uses this

  distance to index the results. \\

They both use a file called 'list' to perform the analysis,  which has

the following structure:  each line contains the type of the value you

want to follow, it can be:

\begin{itemize}

\item[b]  for a Bond distance

\item[a] for an angle

\item[d] for a dihedral

\item[c] to create a center of mass

\end{itemize}

This descriptor is followed by the number of the atoms involved.  The

exception, is the \texttt{c} command that is followed by the number of

atoms used to define this center of mass, and then the index of the atoms.

A typical file can be:

\begin{verbatim}

 b  1  2

 b 2  3

 a 1 2 3

 c  2 1 2   <- create a new atom located at the middle of the 1-2 bond.

\end{verbatim}

The cartesian geometries have to follow the XMol format. If the

comment line contains \texttt{E=} then \texttt{xyz2scan} and

\texttt{xyz2path} will read the following number and take it as the

image energy.

\end{itemize}

\end{document}