Hands-on Tutorials¶
QUICK home directory contains some test runs and examples. These can be found inside the examples directory. For this tutorial we will use water as an example. The structure of a typical QUICK input is as follows.
HF BASIS=cc-pVDZ CUTOFF=1.0d-10 DENSERMS=1.0d-6 ENERGY <== Job command
<== Empty line
O -0.06756756 -0.31531531 0.00000000 |
H 0.89243244 -0.31531531 0.00000000 |<== Coordinates
H -0.38802215 0.58962052 0.00000000 |
Depending on the installation, you may run QUICK as folllows.
Serial version:
./quick input.in
MPI version:
mpirun -np num_cores ./quick.MPI input.in
where num_cores is the number of cores that you wish to use.
CUDA version:
./quick.cuda input.in
Now assume that we have successfully installed CUDA version and set the basis set path.
1. HF/DFT single point energy calculation¶
The typical job command line for HF/DFT is as follows. Note that all keywords are seperated by a single space.
HF BASIS=cc-pVDZ CUTOFF=1.0d-10 DENSERMS=1.0d-6 ENERGY
^ ^ ^ ^ ^
| | | | |
| | | | Compute energy
| | | Density matrix cutoff
| | Integral cutoff
| Basis set
Hamiltonian
For DFT, we replace HF hamiltonian with DFT and specify the functional name separated by a space. For example, we may ask for a B3LYP energy calculation with the following command line.
DFT B3LYP BASIS=cc-pVDZ CUTOFF=1.0d-10 DENSERMS=1.0d-6 ENERGY
^ ^ ^ ^ ^ ^
| | | | | |
| | | | | Compute energy
| | | | Density matrix cutoff
| | | Integral cutoff
| | Basis set
| Functional
Hamiltonian
Note that in the above command, we are using NATIVE B3LYP functional. If we want to use LIBXC B3LYP functional, the functional command should change as follows.
DFT LIBXC=HYB_GGA_XC_B3LYP BASIS=cc-pVDZ CUTOFF=1.0d-10 DENSERMS=1.0d-6 ENERGY
^ ^ ^ ^ ^ ^
| | | | | |
| | | | | Compute energy
| | | | Density matrix cutoff
| | | Integral cutoff
| | Basis set
| Functional
Hamiltonian
It is also possible to ask for exchange and correlation LIBXC functionals seperately. For instance, if we use BLYP, the functional command changes as follows.
DFT LIBXC=GGA_X_B88,GGA_C_LYP BASIS=cc-pVDZ CUTOFF=1.0d-10 DENSERMS=1.0d-6 ENERGY
^ ^ ^ ^ ^ ^
| | | | | |
| | | | | Compute energy
| | | | Density matrix cutoff
| | | Integral cutoff
| | Basis set
| Functionals (Functional_1, Functional_2 seperated by a comma)
Hamiltonian
Note that currently, QUICK cannot handle more than two functionals at a time.
We now proceed with HF single point energy calculation for water molecule. Our input file, water.in is as follows.
HF BASIS=cc-pVDZ CUTOFF=1.0d-10 DENSERMS=1.0d-6 ENERGY
O -0.06756756 -0.31531531 0.00000000
H 0.89243244 -0.31531531 0.00000000
H -0.38802215 0.58962052 0.00000000
Executing QUICK will give us .out and additionally, we save the terminal output into a .run.log file.
./quick.cuda water.in > water.run.log
The content in water.run.log is not useful for users. If debug flags are enabled, information useful for developers can be found here.
The information reported in .out (from now on output) file are as follows. In the beginning of output file, we can find information about job card and the GPU used for the calculation. The next section reports information from SAD initial guess. This will be followed by some information about our molecule such as input geometry, basis function information and etc.
=========== Molecule Input ==========
| TOTAL MOLECULAR CHARGE = 0 MULTIPLICITY = 1
| TOTOAL ATOM NUMBER = 3 NUMBER OF ATOM TYPES = 2
| NUMBER OF HYDROGEN ATOM = 2 NUMBER OF NON-HYDROGEN ATOM = 1
| NUMBER OF ELECTRONS = 10
-- INPUT GEOMETRY -- :
O -0.0676 -0.3153 0.0000
H 0.8924 -0.3153 0.0000
H -0.3880 0.5896 0.0000
-- DISTANCE MATRIX -- :
1 2 3
1 0.00000
2 1.81414 0.00000
3 1.81414 2.96300 0.00000
============== BASIS INFOS ==============
| BASIS FUNCTIONS = 25
| NSHELL = 12 NPRIM = 32
| JSHELL = 12 JBASIS = 32
Next we find information on SCF iterations.
------------------------------------------------------------------------------------------------------------------------
NCYC ENERGY DELTA_E SCF_TIME DII_CYC DII_TIME O_TIME DIAG_TIME MAX_ERR RMS_CHG MAX_CHG
------------------------------------------------------------------------------------------------------------------------
1 -76.056050700 ------ 0.307 1 0.29 0.02 0.00 0.1775E+01 0.5918E-01 0.3593E+00
2 -75.980565869 -.754848E-01 0.010 2 0.00 0.01 0.00 0.2376E+00 0.1554E-01 0.1750E+00
3 -76.017433601 0.368677E-01 0.010 3 0.00 0.01 0.00 0.1050E+00 0.4979E-02 0.6042E-01
4 -76.025458827 0.802523E-02 0.010 4 0.00 0.01 0.00 0.2584E-01 0.1707E-02 0.1991E-01
5 -76.026128208 0.669381E-03 0.010 5 0.00 0.01 0.00 0.4594E-02 0.7144E-03 0.5988E-02
6 -76.026196776 0.685678E-04 0.010 6 0.00 0.01 0.00 0.9251E-03 0.1740E-03 0.1141E-02
7 -76.026199618 0.284200E-05 0.010 7 0.00 0.01 0.00 0.1452E-03 0.3909E-04 0.2857E-03
8 -76.026199744 0.126052E-06 0.010 8 0.00 0.01 0.00 0.3826E-04 0.7945E-05 0.7236E-04
9 -76.026199750 0.583184E-08 0.010 9 0.00 0.01 0.00 0.9753E-05 0.2119E-05 0.1871E-04
10 -76.026199750 0.388203E-09 0.011 10 0.00 0.01 0.00 0.2026E-05 0.4872E-06 0.4202E-05
------------------------------------------------------------------------------------------------------------------------
REACH CONVERGENCE AFTER 10 CYLCES
MAX ERROR = 0.202570E-05 RMS CHANGE = 0.487164E-06 MAX CHANGE = 0.420193E-05
-----------------------------------------------
This is followed by electronic, nuclear and total energies.
ELECTRONIC ENERGY = -85.183315734
CORE_CORE REPULSION = 9.157115983
TOTAL ENERGY = -76.026199750
Finally, we find timing information of the calculation.
The output of a DFT energy calculation is very similar to that of HF, however, with a couple of exceptions. In the job information section, we can find information about density functional being used. If the functional is from LIBXC library, related information will show up. Furthermore, timing section at the very end of output will contain time for performing grid operations (i.e. grid generation, octree run and etc.) and time for computing exchange correlation energy and matrix elements of potential.
2. HF/DFT gradient calculation¶
The HF/DFT gradient calculation input is the same that we used for single point energy, except the ENERGY keyword is now replaced with GRADIENT. Our water example input is now modified as follows.
HF BASIS=cc-pVDZ CUTOFF=1.0d-10 DENSERMS=1.0d-6 GRADIENT
O -0.06756756 -0.31531531 0.00000000
H 0.89243244 -0.31531531 0.00000000
H -0.38802215 0.58962052 0.00000000
In the calculation output, we can find the nuclear gradients immediately after the SCF cycles and energy information but before the timings. For above example the printed gradients will be as follows.
ANALYTICAL GRADIENT:
----------------------------------------
COORDINATE XYZ GRADIENT
----------------------------------------
1X -0.0675675652 0.0126073406
1Y -0.3153153341 0.0180535055
1Z 0.0000000000 -0.0000000303
2X 0.8924325081 -0.0049459616
2Y -0.3153153341 -0.0099345180
2Z 0.0000000000 0.0000000419
3X -0.3880221796 -0.0076370422
3Y 0.5896205650 -0.0080873988
3Z 0.0000000000 -0.0000000115
----------------------------------------
Note that current version of QUICK prints only the total gradients. Therefore, HF and DFT gradient printout remains the same. Finally, in the timings section we can find gradient timings for 1e, 2e and exchange correlation gradient calculations.
3. HF/DFT geometry optimization calculation¶
For HF/DFT geometry optimizations, we should specify OPTIMIZE keyword in QUICK input. For instance, the geometry optimization input for our water molecule would be:
HF BASIS=cc-pVDZ CUTOFF=1.0d-10 DENSERMS=1.0d-6 OPTIMIZE
O -0.06756756 -0.31531531 0.00000000
H 0.89243244 -0.31531531 0.00000000
H -0.38802215 0.58962052 0.00000000
QUICK geometry optimization output will contain information of SCF, gradient and cartesian coordinates for each optimization step. As in the gradient calculation, the analytical gradients will be printed out immediately after the SCF information.
ANALYTICAL GRADIENT:
----------------------------------------------------------------------------
VARIBLES OLD_X OLD_GRAD NEW_GRAD NEW_X
----------------------------------------------------------------------------
1X -0.0876252350 0.0000197070 0.0000066787 -0.0876309328
1Y -0.3438974485 0.0000305134 0.0000178110 -0.3439209444
1Z 0.0000002515 -0.0000000336 -0.0000000335 0.0000002992
2X 0.8793809511 -0.0000260952 0.0000050220 0.8793829892
2Y -0.2849281308 0.0000367010 0.0000272289 -0.2849766745
2Z -0.0000003515 0.0000000484 0.0000000483 -0.0000004202
3X -0.3550174280 0.0000168026 -0.0000012871 -0.3550285769
3Y 0.5874160803 -0.0000028098 0.0000193489 0.5873964801
3Z 0.0000001020 -0.0000000153 -0.0000000153 0.0000001238
----------------------------------------------------------------------------
Next we find information essential for the convergence of geometry optimization.
OPTIMIZATION STATISTICS:
ENERGY CHANGE = -0.9827189729E-08 (REQUEST= 0.10000E-05)
MAXIMUM GEOMETRY CHANGE = 0.4854368017E-04 (REQUEST= 0.18000E-02)
GEOMETRY CHANGE RMS = 0.1958922994E-04 (REQUEST= 0.12000E-02)
GRADIENT NORM = 0.1292934393E-04 (REQUEST= 0.30000E-03)
The cartesian coordinates of the molecular structures pertaining to each optimization step are printed next.
OPTIMIZED GEOMETRY IN CARTESIAN
ELEMENT X Y Z
O -0.0876 -0.3439 0.0000
H 0.8794 -0.2850 -0.0000
H -0.3550 0.5874 0.0000
We can also find the energy of the minimum structure at the end of output but before the timings are printed out.
Last updated by Madu Manathunga on 03/25/2020.