Copy the template.input file and rename it so that the base name matches the name of the coordinate (.car) file. The file must have the .input extension to be recognized.
Use vi editor to examine the input file and change the options. The input file consists of a list of keywords followed by values. Optional keywords are commented out using a number sign (#) at the beginning of a line. A list of the most important keywords is given below.
The value of Calculate tells DMol which type of calculation you want to perform.
Calculate value_keyword
energy
(single-point energy)A single-point energy task computes the SCF solution for the input geometry. By default, the least computationally demanding options are chosen, which sets up a calculation suitable for studying electronic structure (molecular orbitals, dipole moment, etc.).
optimization
(geometry optimization)A geometry optimization task optimizes your model to a stable geometry, that is, it locates a minimum in the potential energy surface.
frequency
(vibrational frequency)To control aspects of how frequencies are computed use the following keywords in the input file:
Vibration_steps
Vibration_project
Vibration_restart
The ability to restart a vibrational frequency calculation is a particular advantage of DMol3. Frequency calculations are time consuming jobs and they sometimes exceed CPU time limits. If the job finishes prematurely simply change the Vibration_restart flag from "off" to "on" and resubmit the job.
optimization_frequency
This option combines the optimization and frequency steps into a single job.
ts_search
The transition-state search is done with the specified input parameters, including constraints (if specified), to produce a transition-state structure. Then this structure is used as the starting point for a frequency calculation.
gradient
To compute the energy and gradient for a given geometry (without optimization), followed by any requested properties choose this option.
Molecular_Dynamics
To propagate the structure using Car-Parinello molecular dynamics based on DFT methods choose this option. Additional menu items are listed under the MD keywords listed in the input file.
Simulated_Annealing
To perform various combinations of molecular dynamics protocols in series this option is used. You may perform melting, damping, quenching, constant energy and constant temperature molecular dynamics protocols.
The Functional keyword specifies the local correlation and gradient-corrected functionals for exchange and correlation.
Functional value_keyword
|
command keyword |
value keywords |
default |
meaning |
|
Functional |
pwc |
pwc |
Perdew-Wang (1992) functional. |
|
vwn |
|
Vosko-Wilk-Nusair (1980) functional. |
|
|
jmw |
|
Hedin-Lundqvist/Janak-Morruzi-Williams (1978) functional. |
|
|
ks |
|
Kohn-Sham (1965) exchange-only functional. |
|
|
gga (pw91) or p91 or GGA |
|
Perdew-Wang (1991) functional. |
|
|
bp |
|
pw91 and b88 functionals. |
|
|
blyp |
|
lyp and b88 functionals. |
|
|
vwn-bp |
|
bp functional with the local correlation replaced by vwn functional. |
|
|
|
bop |
|
Becke-Tsuneda-Hirao gradient corrected exchange-correlation functional. |
The extension of the Hedin-Lundqvist functional to spin-polarized systems is the von Barth-Hedin functional. DMol3 does not use the original BH parameters, instead using the ones used by Janak, Moruzzi, and Williams (JMW) from their original work on metals.
The Vosko-Wilk-Nusair (VWN) functional is the most popular LSD correlation potential. It uses a fit to accurate numerical results (by Ceperly and Alder) of a uniform electron gas. Ceperley and Alder did quantum Monte Carlo calculations on a uniform electron gas at low- and high-spin limits for several electron densities. VWN used the Pade interpolation procedure to fit the CA results for both the para and ferro states and for low and high densities. DMol3 uses the best VWN (so called "Fit") parameters.
The Perdew-Wang (PWC) functional is a recent parametrization of the Ceperley and Alder data, which corrects some VWN problems with fitting. This is the default for DMol3 calculations.
The LSD approximation can be used to quite accurately predict structures, vibrations, and relative energies of covalent systems. However, bond energies are seriously overestimated. The local DFT should not be used for systems with weak bonds, such as hydrogen bonds. These problems with the LSD method can be corrected to a large extent by using the so-called gradient-corrected (or nonlocal) functionals.
DMol3 supports several nonlocal exchange and correlation functionals. The most popular, the Becke exchange functional (B88) is used in conjunction with the Perdew-Wang correlation functional (BP) or the Lee-Yang-Parr correlation functional (BLYP). The latest, so-called generalized gradient corrected (GGA) functional, by Perdew and Wang (P91) was derived by considering low- and high-density regimes and by enforcing various summation rules.
Although the NLSD methods are significantly better then the LSD method, particularly in studying chemical reactions, the NLSD methods may still lead to reaction barriers that are too low.
The VWN-BP functional is recommended for COSMO and COSMO-RS studies.
DMol3 4.2 MatSci includes a new gradient corrected Hamiltonian, named BOP. This Hamiltonian consists of two parts: exchange functional of Becke (1988a) and correlation functional of Tsuneda and Hirao (1997). The correlation part is a much improved Colle-Salvetti-type correlation functional for both opposite spin and parallel spin correlations and it obeys all the necessary conditions of the exact correlation functional. It was found that the new BOP functional improves dramatically accuracy of DFT calculations for organic-type systems.
The Functional_Post_LDA keyword allows for calculation of the energy using a gradient-corrected functional at the end of the calculation.
Functional_Post_LDA value_keyword
|
command keyword |
value keywords |
default |
meaning |
|
Functional_Post_LDA |
off |
off |
Do not perform NLDA calculations after LDA calculations. |
|
BP |
|
Use the BP NLDA functional. |
|
|
BLYP |
|
Use the BLYP NLDA functional. |
|
|
GGA or P91 |
|
Use the GGA NLDA functional. |
|
|
VWN-BP |
|
BP functional with the local correlation replaced with VWN functional. |
A typical use of this command would be to carry out a geometry optimization at the local DFT level, followed by calculation of nonlocal DFT energies using the local DFT-optimized structure.
The Harris keyword activates the use of Harris functional approximation during the calculation. This option significantly reduces the CPU time, since practically only one SCF iteration is done.
Harris value ...
|
command keyword |
value keywords |
default |
meaning |
|
Harris |
off |
off |
Do not use Harris functional. |
|
on |
|
Use Harris functional. |
DMol3 supports several types of representations for the atomic core of atoms.
No core representation. Perform all-electron calculation.
Use relativistic Effective Core Potentials. This reduces the computational cost by collecting core electrons into a single analytical representation. Currently ECPs are available for elements from Z=21 (Sc) to 103 (Lr).
Use all-electron with relativity. This does include relativistic effects that are important for heavy elements while retaining more accuracy than an ECP calculation.
A number of basis sets are listed in the template input file. Using the basis keyword. The parameters in the basis set will be listed in the .inatom file.
If you wish to specify a user-defined basis set you will need to set the Atom_Calculation flag to "inatom" or simply remove the "#"-sign that comments out that line.
Basis value_keyword
[value_number ...
blank line]
|
command keyword |
value keywords |
default |
meaning |
|
Basis |
min |
|
Use minimal basis set; 1 atomic orbital for each orbital that is occupied in the free atom. |
|
dn |
|
Use double-numeric quality basis set; approximately 2 atomic orbitals for each 1 occupied in free atom. |
|
|
dnp |
|
Use DN basis with polarization functions, i.e., functions with angular momentum one higher than that of highest occupied orbital in free atom 1. |
|
|
dnd |
dnd |
Double-numerical + d-DNP basis except that no p functions are used on hydrogen. |
|
|
mixed |
|
Allows you to customize the basis set, starting on the following line. |
|
|
extended |
|
Use additional polarization functions as calculated by the atomic program. |
|
|
all |
|
Use all the orbitals calculated by the atomic program. |
|
|
user |
|
Select your own orbital basis set from the orbitals calculated by the atomic program. |
The template file species a DNP basis set.
The Atom_Rcut keyword allows you to specify the use of real-space cutoffs in the calculation of atomic basis sets.
Atom_Rcut value
|
command keyword |
values |
default |
meaning |
|
Atom_Rcut |
real > 0 |
5.5 |
The basis set cutoff in angstroms. |
You can input Atom_Rcut in Bohrs by specifying:
Atom_Rcut real>0 AU
For Harris calculations, you may use an even smaller cutoff of 4.5 Å. A smaller cutoff improves computational performance but may diminish the accuracy of calculations.
The Integration_Grid keyword controls the selection of mesh points for the numerical integration procedure used in the evaluation of the matrix elements. Since the functions in these equations are given numerically, the choice of mesh points can have a significant impact on the final results.
Integration_Grid value_keyword1 [value_keyword2]
|
Command keyword |
value keywords |
default |
value_keyword2 |
meaning |
|
Integration_Grid |
xfine |
medium |
adaptive |
The geometry independent grid is the default. The adaptive grid can still be used by setting value_keyword2 as adaptive. If the user option is only available for the adaptive grid. |
A value of medium appears to give a reasonable compromise between computational expense and numerical precision.
A value of fine is used whenever difficulties are found with the medium mesh. Such problems can include discontinuous energies (i.e., bumpy rather than smooth potential energy surfaces are observed) and failure of a geometry optimization.
A value of xfine results in a nearly saturated mesh. This option is generally used only for benchmark calculations, not for day-to-day production runs.
A value of coarse results in lower computational cost and a significant reduction in the quality of the calculation. The results are at least qualitatively correct.
A value of xcoarse represents a serious compromise in the quantitative aspects of the calculation. It is useful for:
A value of user allows you complete flexibility in specifying the integration mesh. Six parameters are required to completely specify the integration mesh. The flags xcoarse, coarse, medium, fine, and xfine select predetermined values for them. The parameters are called ipa, iomax, iomin, thres, rmaxp, sp. The mesh is specified in spherical coordinates around each atomic center. The values of rmaxp and sp control the radial distribution of points, and iomax, iomin, and thres control the angular sampling.
Radial points are taken in shells from the nucleus out to a distance of rmaxp (default = 10.0 au). The number of shells is a function of the nuclear charge, so that heavier elements have a greater number of points. The actual number is 14
X (Z + 2)1/3, where Z is the atomic number. The value of sp may be used to increase or decrease this number, that is, sp = 2.0 results in twice as many radial points.Angular sampling points are generated on each shell using a spherical harmonic function. The initial number of angular sampling points is 14, generated with a harmonic function with l = 5. The number of angular points must be increased at larger radii in order to maintain the quality of the numerical integration. The value of thres specifies the requested numerical precision (default = 0.0001), and the number of angular points is increased to maintain the precision. The minimum value of the harmonic generator is given by iomin (default = 1) and the maximum allowed value by iomax (default = 4). The value cannot increase beyond iomax, even if numerical precision is compromised.
The following table gives the angular momentum of the spherical harmonic function used to generate the points on a radial shell and the number of such points:
|
iomin (or iomax) |
l |
# points |
|
1 |
5 |
14 |
|
2 |
7 |
26 |
|
3 |
11 |
50 |
|
4 |
17 |
110 |
|
5 |
23 |
194 |
|
6 |
29 |
302 |
|
7 |
35 |
434 |
|
8 |
41 |
590 |
The value of ipa (default = 6) controls the partition function used to improve the convergence of the numerical integration.
The default values for each of these parameters that are selected by the flags xcoarse, coarse, medium, fine, and xfine:
|
Integration_Grid value |
ipa |
iomax |
iomin |
thres |
rmaxp |
sp |
|
xcoarse |
6 |
3 |
1 |
0.01 |
10.0 |
1.0 |
|
coarse |
6 |
4 |
1 |
0.001 |
10.0 |
1.0 |
|
medium |
6 |
6 |
1 |
0.0001 |
10.0 |
1.0 |
|
fine |
6 |
6 |
1 |
0.00001 |
12.0 |
1.2 |
|
xfine |
6 |
7 |
1 |
0.000001 |
15.0 |
1.5 |
Charge
Regardless of the type of calculation you want to perform, if your (non-periodic) model is not neutral you need to inform DMol3 of its charge by entering a value after the Charge keyword. Enter 0 for a neutral molecule, 1 for a singly charged cation, -1 for a singly charged anion etc. For periodic models, the Charge must be zero.
Restricted or unrestricted spin polarization
Likewise, if you want to consider a particular spin multiplicity, use the Spin Polarization keyword to specify a restricted or unrestricted spin calculation.
To specify properties comment out properties such as Mulliken and Hirshfeld Analysis and HOMO and LUMO grids that are listed as keywords in the input file.
To change the default values for the parameters that control SCF convergence, change the values afer the SCF Options... keywords in the input file.
The Occupation keyword specifies orbital occupations and improves SCF convergence.
Occupation value_keyword
[ range value_number ...]
|
Command keyword |
value keywords |
default |
meaning |
|
Occupation |
Fermi |
Fermi |
Allow DMol3 to determine optimal orbital occupation that yields lowest energy. |
|
Thermal range {real} |
0.02 |
Smear electrons over several orbitals using finite-temperature Fermi function, which can be input with the range parameter. |
|
|
Level_Shift range {real} |
0.25 |
Separate occupied and virtual orbitals by shifting up virtual orbitals by range a.u. |
|
|
Freeze range {integer} |
5 |
Optimize orbital occupancy for all iterations up to range, then fix occupancy. |
|
|
Fixed |
|
Fix occupancy according to information from .occup file. |
|
|
Level_Track |
|
Occupy those states with low-lying eigenvectors that have maximal projection onto the occupied subspace of the previous iteration. |
DMol3, by default, finds the occupations that yield the lowest energy. This means that electrons occupy orbitals with the lowest orbital energies. The occupation numbers are integers.
However, there may be a need to use a fractional occupancy, which effectively mixes some virtual orbitals into the occupied space. This is used when the HOMO-LUMO gap is small and there is significant density of states in the vicinity of the Fermi level. In such situations, you can achieve SCF convergence by using smearing techniques, shifting up the virtual orbitals, or fixing occupancy. Typically, you can encounter problems with SCF convergence for high-symmetry, open-shell, or metallic systems.
The Thermal option uses a finite-temperature Fermi function to achieve fractional occupancy. It also implements an electronic entropy term, which is necessary for accurate derivatives of geometry-dependent orbital populations. The fractional occupancy pattern depends on the temperature in thermal smearing. If smearing of electrons is needed, the Thermal option is recommended.
If there should be difficulties in achieving SCF convergence (e.g., in metallic systems due to small HOMO-LUMO gap), we recommend the use of Thermal occupation instead of Fermi (default). As shown here, an additional keyword is required in order to use Thermal occupation in periodic calculations.
DMol3 allows you to include the effect of an applied uniform electric field on molecular properties.
Externally applied electric fields can be calculated using the Electric_Field keyword. The magnitude of the applied field is given in atomic units and is listed in the x, y and z directions, respectively. The molecule must be oriented appropriate in order for meaningful results to be obtained.
A reaction field that corresponds to solvation a molecule by a dielectric continuum can be calculated using the Cosmo keyword. The effective dielectric constant can be specified using the Cosmo_Dielectric keyword.
Options to control the amount of output produced by DMol3 are found under print options at the bottom of the input file.