linearly dependent basis set
EIGS (block 3): print out the eigenvalues of the overlap matrix (S)
This is a test run
Basis set calibration
A possible strategy for calibrating basis sets for metals is then as follows:
1) Start from a good atomic basis set from one of the libraries mentioned above.
2) Keep all the exponents with values larger than 1.
3) Start with some guess exponents, e.g. sp-shells with values 1., 0.3, 0.1. They should be separated by factors in the range of 2-4 to avoid linear dependence and to have an approximatively complete basis set. Then, add at least one d-shell for 1st row transition metals, e.g. with a value of 0.4. For 2nd and 3rd row transition metals you should have one more diffuse d-function, e.g. try 0.15. This is usually already a very good guess.
4) Reoptimize the exponents: determine the tight sp exponent which has the lowest energy while keeping the other exponents fixed, then the second sp exponent, then the third sp exponent, then the d exponent. After this start again with the tight sp exponent. Repeat this until the energy does not change any more.
When optimising a Gaussian basis set for periodic systems, it is good to minimise the total energy. But a low total energy does not guarantee a good basis set. One should also compute the band structure and compare with the literature. If that does not match, then it is usually a hint that there are basis set deficiencies, typically missing diffuse functions. Obviously, the same functional should be compared (or functionals which perform similar), the literature should be reliable, and so on (geometry, relativity or not, etc).
Basis set superposition error (BSSE) correction
Counterpoise correction
binding energy of AB
ΔE = E(AB)AB - E(A)a - E(B)b
To get the BSSE estimation from the counterpoise correction method, we need four additional single-point calculations with the geometry of compound AB denoted by †.
ΔECP = E(A)†ab + E(B)†ab - E(A)†a - E(B)†b
where E(A)†ab includes ghost orbitals of B and E(B)†ab includes ghost orbitals of A.
Then, Ebind = ΔE - ΔECP
Cohesive energy
Adhesion energy of MgO single layer on Ag
Ag위에 MgO single layer의 binding energy를 구한다고 하자.
Eads = E(MgO/Ag) - E(MgO) - E(Ag) = -0.475 eV
Eads(BSSE) = E(MgO/Ag) - E(MgO, BSSE) - E(Ag, BSSE) = -0.290 eV
이때 E(MgO,BSSE)는 MgO/Ag 구조에서 Ag를 GHOST atom으로 처리하고, E(Ag,BSSE)는 MgO/Ag 구조에서 MgO를 GHOST atom으로 처리.
gCP scheme
atom-pair wise counterpoise corrections directely added to the KS-DFT energies
!Note that the gCP scheme has been carefully tested for molecular crystals. Applications to inorganic crystals and metals are less well tested and should be treated carefully.
The gCP has been developed for a range of standard basis sets, including MINIS, def2-SV(P), def2-SVP, def-TZVP, def2-TZVP, 6-31G*, pob-TZVP.
Memory
“segmentation fault occurred” 에러메세지가 나타나면 메모리 부족을 의심해 볼 수 있다. qstat -f jobid 에서 vmem을 체크해보고 메모리가 부족하면 늘릴 것.
Band structure
One-electron properties
SCF계산이 끝나면 wavefunction file (fort.9)을 저장한다. fort.9 (unformatted) 가 없을경우 fort.98 (formatted)를 fort.9으로 변환해주는 RDFMWF 를 inputfile (.d3)의 맨 위에 추가해준다.
Bilbao 또는 SeeK-path 으로부터 원하는 k-path 설정한다. Ex) MgO: Fm-3m (225)
##
RESTART CALCULATIONS
geometry 재시작할때 이전 계산에서의 마지막 구조(optcxxx)를 fort.34로 복사한 뒤 EXTERNAL 명령어를 입력하면 자동으로 fort.34를 읽는다.
OPTGEM에 RESTART를 설정하면 구조최적화를 계속해서 진행할 수 있다. OPTINFO.DAT 파일이 필요하다.
input deck의 구조에 유의하자.
Geometry optimization
Diamond
EXTERNAL
OPTGEOM
RESTART
ENDOPT
END
이전 계산의 마지막 density matrix (fort.20)를 읽어들이도록 하자. input block 4에 GUESSP를 추가해 주면 자동으로 읽는다.
Hamiltonian,computational parameters, SCF & C
TOLINTEG
7 7 7 9 30
TOLDEE
7
GUESSP
SHRINK
12 0 24
fortran files
fort.1:
fort.3: reducible overlap matrix
fort.8:
fort.9: The density matrix, constructed from a superposition of atomic densities (unformatted)
fort.98: The density matrix, constructed from a superposition of atomic densities (formatted)
The wave function data are read (when fort.98 is used, the first keyword must be RDFMWF, to convert formatted data in binary data) when running properties, and are not modified.
fort.10: Fock/KS eigenvectors
fort.11: reducible Fock/KS matrix
fort.13: reducible density (and spin) matrix
fort.17
fort.18
fort.19
fort.20: rename fort.9 to fort.20 for reading density matrix for initial guess (GUESSP)
fort.24: Formatted output of Mulliken Population Analysis
fort.25: Formatted output for plotting (it could be Compton profile, electron momentum density, the charge density (and its derivatives, if requested))
fort.28
fort.29
fort.31: electrostatic potential
fort.32: information on the system and the computational parameters
fort.33: coordinates of atoms in the unit cell (input for Xmol, Molden)
fort.34: geometry (in Å) and symmetry operators of the system
fort.38
fort.40
fort.60
fort.61
fort.71
fort.72
fort.82
fort.83
fort.95