Graeme Henkelman Associate Professor henkelman@cm.utexas.edu Office: WEL 3.202A Phone: (512) 471-4179
Research
Saddle point finding methodsThe efficiency of saddle point finding methods are compared on a common test system.
Long timescale dynamicsThe dimer min-mode following method has been combined with the kinetic Monte Carlo algorithm to create an adaptive KMC method.
Catalysis at surfaces Density functional theory is used to determine the mechanisms of catalytic reactions on surfaces.
Nanoparticle catalysisThe electronic structure is calculated for bimetallic nanoparticles, making it possible to optimize the structure and composition of the particles for a desired reaction.
A. Methods for calculating rates and long time scale dynamics The group has developed several methods for calculating rates of slow transitions, such as chemical reactions and diffusion. The methods are based on transition state theory (TST), with or without the harmonic approximation. Within the harmonic approximation, the challenge is to find the relevant saddle point(s) on the energy surface, as well as the harmonic frequencies at the saddle point and at the initial state. In some cases only the initial state is known. Then the 'Minimum-mode following' method is a powerful tool for finding both the mechanism of likely transitions as well as the activation energy. We have, for example, used this to simulate long time scale dynamics in solids. When both the initial and final states are known, the CI-NEB method is the most powerful one we know of. A review sumary of various methods was recently published. We have in particular applied these methods to calculations of crystal growth. Within full TST, the challenge is to find the optimal dividing surface with dimensionality D-1 (where D is the number of degrees of freedom in the system). The optimal surface corresponds to the dividing surface with maximum free energy on the way from the initial state to final state. This represents the tightest bottle neck for the transition. We have developed a method, called OH-TST for finding an optimal hyperplanar dividing surface in a systematic way and a mosaic of several hyperplanes enclosing the intial state. In both the harmonic and full TST calculations, the challenge is to navigate in a high dimensional space. The group has also developed an extension of classical TST to quantum TST, where tunneling can become an important transition mechanism. Here is a more complete list of publications on these topics.
B. Density functional theory calculations The group mainly uses plane wave based DFT to calculate the interaction between atoms in condensed systems. While a parallel DFT program was developed in the group in the early 90s, the rapid development of the methodology and maturity of the field has made it more practical to adopt some of the sophisticated packages now available such as the DACAPO code and the VASP code. Methods for calculating rates (see above) and analysis of electron density developed by the group have been implemented within these packages, see in particular VASP implementations (maintained by former graduate student, Prof. Graeme Henkelman at UT). A particular focus of the DFT calculations in the group is the testing and development of DFT methods for calculating properties of excited states, such as self-trapped excitons in solids, catalysis, and surface diffusion and island formation The development and implementation of orbital dependent functionals in DFT is an ongoing effort in the group.
This is not the official VASP web page. Designed to simulate the properties of systems at the atomic scale, VASP (Vienna Ab-initio Simulation Package) is a software package created, distributed, and maintained by the Hafner Research Group at the University of Vienna.
The code on this site was written by several people who were in or associated with theJónsson group. Development and maintenance are now being coordinated in theHenkelman group at UT Austin.
We have a discussion forum to address issues related to the code and scripts
The following are a set of scripts to perform common tasks to help with VASP calculations, and particularly with transition state finding. The included Vasp.pm perl module contains several simple routines that are used by many of the scripts.
Install by uncompressing this file, and adding the vtstscripts directory to your path.
The scripts are organized into the following categories (portals to them are on the left):
general
file conversion
nudged elastic band
dynamical matrix
dimer
charge density
density of states
Notes:
We recommend that the first line in the POSCAR file contain the element symbols, in the same order as they appear in the POTCAR. This will allow for proper visualization when files are converted to xyz files.
For NEB scripts, there needs to be OUTCAR files for the initial and final states placed in the 00 and NI+1, respectively. Please direct questions about these scripts to the discussion `forum`_.
Perl module that contains various common commands that one might want when deal with VASP POSCAR files. These include reading and writing a POSCAR file, reading and writing a generic vector file, doing dot products and finding magnitudes of vectors and other similar functions
vef.pl
usage: vef.pl
Prints the force and energy at each iteration of a vasp run.
vfin.pl
usage: vfin.pl(outputdirectory)
This script finds the ICHAIN tag from the OUTCAR and cleans up the run directory accordingly. All relevant files (POSCAR, CONTCAR, OUTCAR (zipped), INCAR, KPOINTS, XDATCAR (zipped), CHGCAR and WAVECAR if the are non-empty) are copied to the output directory. In the run directory CONTCARs are moved over POSCARs in preparation for a new run.
Takes initial and final POSCAR files, and linearly interpolates the specified number of images between them. The interpolated files are written to the directories 00 to NI+1, where NI is the number of specified images.
Sets up a dimer run from a NEB run. It is assumed that the configuration is contained in POSCARs, i.e. vfin.pl has been run. If no input argument is given then the dimer is formed by interpolation around the highest point in the exts.dat file. Otherwise it is formed around the input image. Curvature data from the MEP is used for the initial orientation of the DIMER.
Sets up a lanczos run from a NEB run. It is assumed that the configuration is contained in POSCARs, i.e. vfin.pl has been run. If no input argument is given then the run is set up by interpolation around the highest point in the exts.dat file. Otherwise it is set up around the input image. Curvature data from the MEP is used for the initial MODECAR.
nebef.pl
usage: nebef.pl
output: energyandforceofimagesintheneb
nebbarrier.pl
usage: nebbarrier.pl
output: energy,distance,andforcesalongtheneb
Generates the file neb.dat which contains the distance between images, the energy of each image, and the force along the band. This data is used by nebspline.pl to generate a force-based cubic spline along the band.
nebspline.pl
usage: nebspline.pl
output: spline.dat,exts.dat,mep.eps
Reads the file neb.dat and creates the files spline.dat,exts.dat and mep.eps. spline.dat is a set of points that describe the spline fitted to the data in the neb.dat while exts.dat contains the location and energy of all extrema found along the curve and mep.eps is a plot of the MEP path.
nebmovie.pl
usage: nebmovie.pl(flag)
output: movie.xyz
Can be used to generate a movie from standard xyz files, generated either by POSCARs (flag=0) or CONTCARs (flag=1) in every directory.
After a run has finished and wrapped up with vfin.pl, the nebresults.pl can be used to run nebef.pl, nebspline.pl, nebmovie.pl and nebconverge.pl automatically.
nebfreeze.pl
usage: nebfreeze.pl(atom)(listofPOSCARfiles)
output: POSCARfiles,tooriginalfiles
Takes an atom number and a list of POSCAR files and then freezes that atom, as well as shifting the contents of each POSCAR file so that that atom has the same position in each cell. This is useful if you need to give all POSCARs in a NEB calculation the same frozen point.
nebavoid.pl
usage: nebavoid.pldistance
output: POSCARfiles,tooriginalfiles
If atoms are closer than the specified distance, the script pushes these atoms apart. The new geometry is written in the POSCAR file, and the old saved as POSCAR_orig.
Warning: this script does not give a set of equally spaced images.
The values in CHGCAR_sum are (CHGCAR1*fact1+CHGCAR2+fact2). By default, fact1=fact2=1.0, so that the output is the sum of the input charge density files
or : dymmatrix.pl(#DISPLACECAR)(DISPLACECAR1)(DISPLACECAR2)...
or : dymmatrix.pl(OUTCAR1)(OUTCAR2)(OUTCAR3)...
output: mass-scaleddynamicalmatrix(freq.mat)
normalmodefrequencies(freq.dat)
eigenvalues(eigs.dat)
andeigenvectors(modes.dat)
Takes the output from the dynamical matrix calculation and creates the matrix. The DISPLACECARs should only contain those degrees of freedom that were calculated in their corresponding OUTCARs (see dymcmpdisp.pl). The scripts now handles the diagonalization itself via package from CPAN. (It could be a bit slow for large matrices).
Takes as input two POSCAR files, n, the number of atoms to include, and the displacement. It then finds the n atoms that have the largest displacement between the two POSCAR files. The file DISPLACECAR is created, which contains the displacements of each degree of freedom (zero, unless the atom is one of the n atoms, in which case it is the entered displacement). This file is ready to use, then, with the dynamical matrix routine.
Similar to dymseldsp.pl, except it only takes one POSCAR and also needs an atom number as input. It then finds the n atoms closest to the chosen atom and these are the atoms given non-zero displacements in the DISPLACECAR file.
dymcmpdisp.pl
usage: dymcmpdisp.pl(DISPLACECAR1)(DISPLACECAR2)
output: DISPLACECARfile,toSTDOUT
Takes as input two DISPLACECAR files, compares them, and outputs a DISPLACECAR file in which those degrees of freedom that are set in one DISPLACECAR but not the other. Thus, if you use dymseldsp.pl to create a DISPLACECAR with 24 degrees of freedom, and then you want to calculate then next 12 degrees of freedom, you would dymseldsp.pl for 36 degrees of freedom and use compare_disp.pl to extract those 12 which aren’t included in the first DISPLACECAR. You can then run the dynamical matrix calculation on this new file, getting the forces for these 12 new displacements, and then use dymmatrix.pl to combine the OUTCARs from both calculations into one matrix.
Used to fit between two or more matrices together. It takes as input the order of the fit, and then pairs of displacements and matrices. It outputs a matrix of the same order. It requires the Perl modules Math::Matrix and Math::Approx.
Used to create a smaller dynamical matrix from a larger one. If you calculated many degrees of freedom the first time and want to check how the quantity converges versus degrees of freedom, use this to create the smaller matrix. You need the DISPLACECAR for the matrix you have and the DISPLACECAR for the matrix you want.
Reorders a dynamical matrix. You might want to use this if you plan on fitting matrices, but you got at them different ways (for example, one you had by doing all of the degrees of freedom at once and the second you created by doing degrees of freedom a bit at a time... the ordering of the displacements in each matrix will be different). You need the series of DISPLACECARs that were used to create each matrix.
Intended to help analyze the convergence of the dynamical matrices and compare the differences both between different displacements as well as different orders of fits. It takes as input pairs of displacements and their corresponding matrix. The first argument is a flag. If it is zero, then a fit is done with each successive matrix added to the points used to determine the fit. The output tells how much the force constants change as each point is added to the fit. If the flag is greater than zero, then a fit is done with the first n, where flag equals n, matrices, and the difference between the force constants calculated for the other matrices and the fitted matrices is printed. The analysis is done for any modes which have a frequency larger than modevalue.
Takes the configuration file (POSCAR), displacement file (DISPLACECAR) and modes file (modes.dat), which is created by running dymmatrix.pl, and creates a movie for each mode. These xyz movies are saved in the moviefolder (if designated) or the current directory. The frequency of each mode will be written to the xyz files if the freq.dat file is provided. The numimages variable sets the number of frames in each mode movie and the dist variable sets the vibrational amplitude.
Note: if the modes.dat is created by using serveral DISPLACECARs as indicated in dymmatrix.pl, use the concatenated DISPLACECAR file in this script (i.e., cat those DISPLACECAR files in the same order as they were used in dymmatrix.pl).
Split the DOSCAR file into atomic DOS files, (DOS1, DOS2,..,DOSN). The DOSCAR and OUTCAR in the working directory will be used. This bash script was written for LORBIT =10 and = 11 calculations as well as spin (un)restricted. The energy will be referenced to the fermi energy specified in the OUTCAR, (E-Ef).
This script sums up the atomic projected DOS over some group of atoms, and then calculates the center of the specific band using a weighted average. The default is set to calcualted the center for the whole band. However, it can also consider DOS within a user specified range by using the optional “w=” or “e=” flag.
By using “e=emin,emax” flag, only the states in ranger [emin,emax] are considered.
By using “w=” flag, the script finds a half width for the band at half the max height and, based on the number following w= , calculates a weighted average within the limits of that many half widths at half height from the center. In this manner, a band center may be found by weighted average without including noncontributing states.
If no orbital flag is specified, the script analyzes the d-band. If no atom is selected, it analyzes all of them. If no w= tag specified, the center is calculated between 2.5 half widths at half height
Note: the split_dos script should be run first to get the resulting new files labeled as DOS1, DOS2, ..., DOSN,where N is the number of atoms in the unit cell.
doslplot.pl
usage: doslplot.pl
output: ldosplot.eps
The eps file has DOS plot for each of selected atoms (blue lines), and the DOS plot for the all system (red line). If no orbital flag, plot the d-band.If no atom is selected, plot all of them.
Note: Only LORBIT=11 and up to s,p,d bands can be handled with this script.