IDM

go back to Main Page, Computational Resources, Chemistry & More, Computational Codes, VASP

DESCRIPTION

VASP_IDM is our internal abbreviation for a set of extra routines for VASP, developed by Dr. Tomáš Bučko (Univ.Vienna), which is useful for setting up calculations on transition state search.

Using the corresponding VASP binary loaded with the module 5.2_IDM, the improved dimer method (IDM) (IBRION=44) and intrinsic reaction coordinate (IRC) search (IBRION=40) become available.

In the following, the (currently not published elsewhere) manual supplied by the developer is presented.

Improved Dimer Method (IDM)

The current implementation does not support lattice optimisation and can be used only for atomic relaxations (ISIF=2).

Flags and parameters

IBRION=44	invokes relaxation with the dimer method
FINDIFF=1|2	forward (1) of central (2) differences formula for numerical differentiation
DIMER DIST=0.01	step for numerical differentiation (Å)
STEP SIZE=0.01	trial step size for energy minimisation(Å)
STEP MAX=0.1	maximal step size for energy minimisation (Å)
MINROT=0.01	minimal rotation of dimer (rad.)

The method is described in detail in Ref.(2). The only input which has to be defined by user is a 3N-dimensional vector defining direction of negative curvature on potential energy hypersurface. It must be specified in POSCAR on place of ionic velocities (see documentation for the POSCAR file). Note that this vector is automatically normalised so it is the only direction that really matters. Output including maximal gradient in the current step, curvature along the dimer direction and the angle through which the dimer is rotated is written in OUTCAR after key words "DIMER METHOD".

(1) G. Henkelman and H. Jónsson. A dimer method for finding saddle points on high dimensional potential surfaces using only first derivatives. J. Chem. Phys., 111(15):7010–7022, 1999. [1]

(2) A. Heyden, A.T. Bell, and F.J. Keil. Efficient methods for finding transition states in chemical reactions: Comparison of improved dimer method and partitioned rational function optimization method. J. Chem. Phys., 123(22):224101, 2005. [2]

Intrinsic Reaction Cooordinate (IRC)

The current implementation does not support lattice optimization and can be used only for atomic relaxations (ISIF=2).

Flags and parameters

IBRION=40	perform IRC search using the DVV method
DVVEHISTORY=5	number of subsequent increasing energy steps taken before simulation is terminated
DVVDELTA0=1.5e-3	the error tolerance Δ0
DVVVNORM0=0.01	magnitude of the velocity vector v0
DVVMINPOTIM=0.025	lower limit for the time step Δt
DVVMAXPOTIM=3.0	upper limit for the time step Δt
DVVMINUS=.FALSE.|.TRUE.	negative direction of the initial velocity vector

The method is described in detail in Ref.(3). Velocity verlet algorithm is used to drive system from (well relaxed!!!) saddle point to nearby minimum along intrinsic reaction coordinate (IRC):

x_i = x_i-1 + v_i-1Δt_i + ½a_i-1Δt_i²

v_i = v_i-1 + ½(a_i-1+a_i)Δt_i .

Direction of atomic velocities (i.e. vector of negative curvature) must be defined in POSCAR. Note that if DVVMINUS is set to =.FALSE., oposite direction of vector defined in POSCAR is used to initiate atomic velocities. In order to keep intermediate structures close to IRC, atomic velocities are damped such as the norm of velocity vector is constant:

|v_i| = v₀ .

The size of integration step is varied during the simulation such as to ensure optimal performance:

Δt_i+1 = Δt_i|Δ₀/Δ_i |^1/3 ,

Δt_i is constrained to interval DVVMINPOTIM < Δt_i < DVVMAXPOTIM. Output (in particular value of intrinsic reaction coordinate (IRC) and corresponding potential energy) is written into OUTCAR file after key words "DAMPED VELOCITY VERLET ALGORITHM:"

(3) H.P. Hratchian and H.B. Schlegel. Following reaction pathways using a damped classical trajectory algorithm. J. Phys. Chem. A, 106(1):165–169, 2002. [3]

Hands-on Guide for IDM+IRC

Finding a transition state (TS) using the improved dimer method (IDM)

1. Suggest the geometry of the TS. Do not care much about the structure of initial state (IS) and final state (FS) of the reaction as it is for (CI-)NEB calculations, simply stretch out bond(s) to be broken or shorten bond(s) to be formed, by hand in a visualizer.

Of course, you can create a reasonable guess using tools like nebmake.pl (VTST tools) from IS and FS, or resubmit a TS resulted from (CI-)NEB calculation. :)

2. Run a frequency job for the TS guess, i.e. (IBRION=5, NFREE=2, very small POTIM). Hopefully you have a "hard" imaginary vibration mode (no matter that there are some soft ones). Extract the displacement vector of the imaginary mode into a file DIMERDIRECTION, e.g. using get-dimer-dir.sh script.

3 Submit the IDM job. In comparison to a usual relaxation run (INCAR+KPOINTS+POSCAR+POTCAR), an additional input file for the module VASP/5.2_IDM is the previously created DIMERDIRECTION file. In the INCAR file, set the NSW-tag to a very high number such as 4000 or so, the calculation needs a lot of ionic steps. The most important keyword switching on the IDM method is IBRION=44 tag.

4 Check the TS you've obtained, i.e. run a frequency job and check that you have 1 "hard" imaginary vibration mode. Well done!

Afterwards, you may want to find the corresponding IS and TS (incl. full energy profile) using the IRC method (vide infra).

Finding the initial state (IS) and final state (FS) using the improved dimer method (IRC)

(Under Construction: The full recipy for IRC calculation will appear here soon!)