BigDFT User Manual
Contents
1. 3 41 The input freq fille 4 XANES calculations Al SADSIRACK A u etn re be eine en Bea ie onen ie end 4 2 Introduction gt 4 wa Ban aa ae 4 4 3 1 Note about the absorber pseudopotential 28 4 4 Description of the output 2 2 CC mm nen 29 4 5 The reversed PAW method 2 222 2m nennen 29 4 6 Test against an exactly solvablecase 2 2 2 22020 30 XC functionals codes 32 A l Native ABINITXC codes 2 2 2 2 Cm onen 32 A 2 libXC functional codes 2 22mm nen 33 Chapter 1 Installing BigDFT The compilation and installation of BigDFT rely on the GNU standard building chain configure make make install BigDFT can be used as an independent program as described in this manual or as a library to be embedded in other softwares like inside ABINIT 1 1 Building the executables 1 1 1 Configure The BigDFT build system is based on standard GNU autotools The end user does not need to have the autotools package installed on his computer the configure script provided in the BigDFT package will create the appropriate Makefile and set all the compilation options like the optimization level the associated libraries to link with and so on After the package has been untarred the sources should be configured to the local architecture of the system Thanks to the autotools it is possible to gen erate several builds from the same source tree I2. e freq_method which determines the method only 1 at the present stage 24 Chapter 4 XANES calculations 4 1 Abstract We have implemented an original procedure for calculating XANES spectra within the BigDFT code Our approach consists firstly in projecting the photoelectron wave function onto the pseudopotential functions basis with the help of a reversed PAW projector 1 and then propagating this initial state by iterative applications ofthe Hamiltonian The Hamiltonian is the one provided by the BigDFT code The obtained spectra therefore correspond exactly to the electronic structure of the self consistent ground state The spectra are built by obtaining at each new application of the Hamiltonian a new coefficient of the spectral decomposition in Chebychev polynomials The method therefore systematically improves the calculated spectra resolution at each step and can be stopped once the required resolution is reached Moreover the Chebychev polynomials have a higher nodes density at the extrema of the spectral range than in the middle For practical applications this makes convergence even faster because the broadening due to lifetime is narrow near the edge and large elsewhere The execution times represent a break through for this kind of calculations and reflect the exceptional features of the BigDFT code For a cluster contained in a 25A side cube the spectra are obtained within half an hour running on a single processo
3. or with A non comment line must contain x y z name giving the 3 coordinates and the name of the atom The coordinates of the atoms are given in the orthonormal basis not in the box basis In case the keyword reduced has been specified the three coordinates x y and z are given in the box basis The keywords can be positionned anywhere after the first three lines The following list summurises all available keywords e reduced All atomic coordinates are given in reduced coordinates with re spect to the box definition e angdeg The box definition contains three distances and three angles instead of the classical six projection values e bohr or bohrd0 The units for the distance are Bohr e angstroem or angstroemd0 The units for the distance are Angstrms 11 e atomic or atomicd0 The units for the distance are Bohr e periodic surface or freeBC The periodicity of the system is either 3D 2D free direction is y or OD Additional informations can be provided after the atom names like for the xyz format 2 3 The input file input dft This file contains all the parameters required for a single wavefunction calculations hgridx hgridy hgridz The grid spacing of the Cartesian grid in Bohr As described above the nodes of this grid serve as the centers for the scaling function wavelet basis Values are in most cases between 3 and 6 crmult frmult Coarse Region Multiplier and Fine Region Multipli
4. Mol Phys 1996 p 1117 dtion the time step for molecular dynamic in atomic time units One atomic time unit is 2 418884e 17 seconds which is the value of Planck s constant in Hartree sec The following lines depend on the choosen value for ionmov if ionmov 6 one line containing the initial temperature in kelvin If nega tive the initial velocities are all zero If positive random speeds are chosen to match the given temperature NOT IMPLEMENTED YET if ionmov gt 7 one line containing two temperatures in kelvin When dif ferent the temperature is linearly change at each geometry step to go from the first value to the second if ionmov 8 one additional line containing the thermostat inertia coeffi cient for Nose Hoover dynamic if ionmov 9 two additional lines containing first the friction coefficient and then a value in Bohr corresponding at a distance where the atoms can bounce on see the ABINIT documentation for further details if ionmov 13 several additional lines containing first the number of ther mostats in the chain of thermostats Then a line with the mass of each ther mostat in the chain And finally a line with two values for the barostat mass depending on optcell value NOT IMPLEMENTED YET 17 2 5 The input file input kpt This file is used to specify a set of k points If this file does not exist only the I point will be used The k point generation relies on the ABINIT
5. the solutions for the pseudo potential and the all electrons case respectively the projector is thus written P Y Wam gt lt Wnta im nlm where A is the number of core orbitals contained in the pseudopotential We show in fig xxx the radial parts W r and W A r for 1 in the case of Iron with a HGH pseudopotential with Zion 16 for n 1 2 xx xxx In figure we show the 29 corresponding pseudo eigenvalues EnJ 1 and the all electrons ones E A 1 1 We can see that the agreement is within XXX for the first XXx eigenvalues and then increase still remaining better than XXX per cent while the agreement between the psuedo wavefunctions and the all electrons ones remains excellent 4 6 Test against an exactly solvable case In progress 30 Bibliography 1 This reversed PAW procedure that we describe and document here looks similar to the one which seems to be used in Taillefumier et al Phys Rev B 66 195107 2002 2 M Blume in Resonant Anomalous X Ray Scattering edited by G Materlik J Sparks and K Fisher Elsevier Amsterdam 1994 p 495 3 A Weiss G Wellein A Alvermann and H Fehske Rev Mod Phys 78 275 2006 31 Appendix A XC functionals codes A 1 Native ABINIT XC codes W oo auDA 12 13 14 15 16 17 20 21 No semi local xc Hartree potential LDA or LSD Teter Pade parametrization S Goedecker M Teter J H tter Phys Rev B54 1703 19
6. written in the file posout_000 xyz frmult 12 crmult 6 hgrid 5 frmult 12 crmult 6 hgrid 3 Figure 2 2 Illustration of the grid and its parameters 2 2 Format of Input Output files for atomic coordinates BigDFT supports atomic files which are of two types The first one is a particular generalisation of the traditional xyz format and can be obtained from this via simple modifications The second one ascii is peculiar of BigDFT and V_Sim Both formats work with the two codes 2 2 1 The xyz format By default the atomic input coordinates are in the file posinp xyz If the same operation e g geometry optimization has to be done for several structures one can also give a list of input files whose names without the xyz extension have to be contained in a file list_posinp The first line in the list_posinp file has to be the number of input files and each consecutive line contains then one filename All the input and output files for atomic coordinates are in the xyz format i e they can be visualized with any standard visualization package and in partic ular with V_Sim The first line contains the number of atoms and then the units which are specified by atomic atomicd0 or Bohr if atomic units are used or by angstroem or angstroemd0O if the angstr m unit is used dO formats 1pe24 17 in Fortran guarantee that not a single bit is lost during write and read of the num bers The s
7. 96 which reproduces Perdew Wang which reproduces Ceperley Alder LDA Perdew Zunger Ceperley Alder no spin polarization LDA old Teter rational polynomial parametrization fit to Ceperley Alder data no spin polarization LDA Wigner functional no spin polarization LDA Hedin Lundgvist functional no spin polarization LDA X alpha functional no spin polarization LDA or LSD Perdew Wang 92 functional LDA or LSD x only part of the Perdew Wang 92 functional LDA or LSD x and RPA correlation part of the Perdew Wang 92 functional GGA Perdew Burke Ernzerhof GGA functional GGA x only part of Perdew Burke Ernzerhof GGA functional GGA potential of van Leeuwen Baerends while for energy Perdew Wang 92 functional GGA revPBE of Y Zhang and W Yang Phys Rev Lett 80 890 1998 GGA RPBE of B Hammer L B Hansen and J K Norskov Phys Rev B 59 7413 1999 GGA HTCH93 of F A Hamprecht A J Cohen D J Tozer N C Handy J Chem Phys 109 6264 1998 GGA HTCH120 of A D Boese N L Doltsinis N C Handy and M Sprik J Chem Phys 112 1670 1998 The usual HCTH functional Fermi Amaldi xc 1 N Hartree energy where N is the number of electrons per cell G 0 is not taken into account however for TDDFT tests No spin polarisation same as 20 except that the xc kernel is the LDA ixc 1 one 32 for TDDFT tests 22 same as 20 except that the xc kernel is the Burke Petersilka Gross hybrid for TDD
8. B98_1b XC_GGA_XC_SB98_Ic XC_GGA_XC_SB98_2a XC_GGA_XC_SB98_2b XC_GGA_XC_SB98_2c XC_HYB_GGA_XC_B3PW91 XC_HYB_GGA_XC_B3LYP XC_HYB_GGA_XC_B3P86 XC_HYB_GGA_XC_O3LYP XC_HYB_GGA_XC_mPW1K XC_HYB_GGA_XC_PBEH XC_HYB_GGA_XC_B97 XC_HYB_GGA_XC_B97_1 XC_HYB_GGA_XC_B97_2 XC_HYB_GGA_XC_X3LYP XC_HYB_GGA_XC_B1WC XC_HYB_GGA_XC_B97_K XC_HYB_GGA_XC_B97_3 XC_HYB_GGA_XC_mPW3PW XC_HYB_GGA_XC_BILYP XC_HYB_GGA_XC_B1PW91 XC_HYB_GGA_XC_mPW1PW XC_HYB_GGA_XC_mPW3LYP XC_HYB_GGA_XC_SB98_la XC_HYB_GGA_XC_SB98_1b XC_HYB_GGA_XC_SB98_lc XC_HYB_GGA_XC_SB98_2a XC_HYB_GGA_XC_SB98_2b XC_HYB_GGA_XC_SB98_2c XC_MGGA_X_LTA XC_MGGA_X_TPSS XC_MGGA_X_MO06L XC_MGGA_X_GVT4 XC_MGGA_X_TAU_HCTH XC_MGGA_X_BR89 XC_MGGA_X_BJ06 XC_MGGA_X_TB09 XC_MGGA_X_RPP09 171 172 173 174 175 176 177 178 179 180 181 401 402 403 404 405 406 407 408 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 201 202 203 204 205 206 207 208 209 35 Boese Martin for Kinetics Becke 97 3 Functionals fitted for water Functionals fitted for water Functionals fitted for water Schmider Becke 98 parameterization la Schmider Becke 98 parameterization 1b Schmider Becke 98 parameterization Ic Schmider Becke 98 parameterization 2a Schmider Becke 98 parameterization 2b Schmider Becke 98 parameterization 2c The original hybrid proposed by Becke The in famous B3LYP Perdew 86 hybrid similar to B3PW91 hybrid using the optx functional mixture of m
9. BigDFT User Manual http inac cea fr L_Sim BigDFT Stefan Goedecker unibas ch http www unibas ch comphys comphys Luigi Genovese esrf fr Damien Caliste cea fr Alessandro Mirone esrf fr Thierry Deutsch cea fr Contents 1 Installing BigDFT 1 1 Building the executables 2 2 22cm 1 1 Configure fs 4 ea era 1 122 Make cc se is an a an aa A ee tO eh ee 1 1 3 Inst ll rs 3 38 ars ae Gee ah en Gees LLA Elean 4 00 23 eae N sea ER e E Ae R 1 2 Building a library 2 2 2 Co oo onen 2 File formats and conventions in BigDFT 2 1 OTHE Basis S tu sune Geos jan bah Ghee Orewa da oe woe abo See aes 2 1 1 One dimensional functions 2 1 2 Wavelet basis sets in three dimensions 2 1 3 Visualizing the simulation grid with V_Sim package 2 2 Format of Input Output files for atomic coordinates 22 1 The xyzformat 000000000 a 2 2 2 The asciiformat 2 3 Theinputfile input dft 2 e onen 2 4 Theinput file input geopt 2 2 2 Cm nn 2 5 Theinput file input kpt siririna CE onen 2 6 Theinput file input mix miaren oo onen 2 7 Theinput file input perf 2 2 2 oo nn 3 Running BigDFT executables 3 1 Estimating the memory usage memguess 3 2 Single point or geometry optimization bigdft 3 3 Doing a path minimization NEB 3 4 Doing frequencies calculations frequencies
10. FT tests 23 GGA of Z Wu and R E Cohen Phys Rev 73 235116 2006 26 GGA HTCH147 of A D Boese N L Doltsinis N C Handy and M Sprik J Chem Phys 112 1670 1998 27 GGA HTCH407 of A D Boese and N C Handy J Chem Phys 114 5497 2001 100 Hartree Fock exchange only A 2 libXC functional codes In the input file input dft you have to specify a code from this table made with a negative sign concatenated with two integers one for exchange part and another one for correlation part XC_LDA_X XC_LDA_C_WIGNER XC_LDA_C_RPA XC_LDA_C_HL XC_LDA_C_GL XC_LDA_C_XALPHA XC_LDA_C_VWN XC_LDA_C_VWN_RPA XC_LDA_C_PZ XC LDA CPZ MOD XC_LDA_C_OB PZ XC_LDA_C_PW XC_LDA_C_PW_ MOD XC_LDA_C_OB_PW XC_LDA_C 2D_AMGB XC_LDA_C_2D_PRM XC_LDA_C_vBH XC_LDA_C_ID_CSC XC_LDA_X_2D XC_LDA_XC_TETER93 XC_LDA_X_1D XC_GGA_X_PBE XC_GGA_X_PBE_R XC_GGA_X_B86 XC_GGA_X_B86_R XC_GGA_X_B86 MGC XC_GGA_X_B88 XC_GGA_X_G96 XC_GGA_X_PW86 OMAN NDNN PWN 101 102 103 104 105 106 107 108 33 Exchange Wigner parametrization Random Phase Approximation Hedin amp Lundqvist Gunnarson amp Lundqvist Slater Xalpha Vosko Wilk amp Nussair Vosko Wilk amp Nussair RPA Perdew amp Zunger Perdew amp Zunger Modified Ortiz amp Ballone PZ Perdew amp Wang Perdew amp Wang Modified Ortiz amp Ballone PW Attacalite et al Pittalis Rasanen amp Marques correlation in 2D von Barth amp Hedin Casula S
11. PW91 and PW91 optimized for kinetics aka PBEO or PBEIPBE Becke 97 Becke 97 1 Becke 97 2 maybe the best hybrid Becke 1 parameter mixture of WC and PBE Boese Martin for Kinetics Becke 97 3 mixture with the mPW functional Becke 1 parameter mixture of B88 and LYP Becke 1 parameter mixture of B88 and PW91 Becke 1 parameter mixture of mPW91 and PW91 mixture of mPW and LYP Schmider Becke 98 parameterization la Schmider Becke 98 parameterization 1b Schmider Becke 98 parameterization Ic Schmider Becke 98 parameterization 2a Schmider Becke 98 parameterization 2b Schmider Becke 98 parameterization 2c Local tau approximation of Ernzerhof amp Scuseria Perdew Tao Staroverov amp Scuseria exchange Zhao Truhlar exchange GVT4 from Van Voorhis and Scuseria exchange part tau HCTH from Boese and Handy Becke Roussel 89 Becke amp Johnson correction to Becke Roussel 89 Tran amp Blaha correction to Becke amp Johnson Rasanen Pittalis and Proetto correction to Becke amp Johnson XC_MGGA_C_TPSS 231 Perdew Tao Staroverov amp Scuseria correlation XC_MGGA_C_VSXC 232 VSxc from Van Voorhis and Scuseria correlation part XC_LCA_OMC 301 Orestes Marcasso amp Capelle XC_LCA_LCH 302 Lee Colwell amp Handy 36
12. alculations By default symmetries are used to update the density in periodic boundary condition calculations and surface boundary conditions also 2 4 The input file input geopt If this file exists BigDFT will do a geometry optimization and the file contains all the required parameters e geopt approach This character string specifies the method used for the geometry optimization VSSD Variable Stepsize Steepest Descent method 15 SDCG A combination of Steepest Descent and Conjugate Gradient LBFGS Limited Memory BFGS BFGS Preconditioned steepest descent with energy feedback A pre conditioning matrix is build up according to the BFGS algorithm The initial Hessian is a diagonal matrix where the diagonal elements are the inverse of the step size This method is usually the most efficient one PBFGS Same as BFGS except that an initial Hessian is obtained from a force field Force field parameters are only available for systems con sisting of H C N O For such systems the method is the most efficient one in general AB6MD The molecular dynamic routines from ABINIT 6 e ncount_cluster_x Maximum number of force evaluations to be used for the geometry optimization e frac_fluct forcemax Convergence criteria for the geometry optimization The geometry optimization stops either if the norm of the individual forces acting on any atom in the system is smaller than forcemax or if the forces get
13. e prefix DIR Specify your installation directory usr local is default An example of compilation using the MKL from Intel instead of basic BLAS and LAPACK installation configur with ext linalg 1mkl_ia32 1lmkl_lapack with ext linalg path L opt intel mk172 1ib 32 prefix home caliste usr FC ifort An other example compiling CUDA parts sources bigdft 1 3 0 dev configure FC mpif90 FCFLAGS 02 assume 2underscores CC icc CXX icc CXXFLAGS 02 I applications cuda 2 2 include CFLAGS 02 I applications cuda 2 2 include with ext linalg L applications intel mkl lib em64t Imkl_scalapack_lp64 lmkl_blacs_intelmpi20_1p64 lmkl_intel_lp64 lImkl_lapack Imkl_sequential Imkl_core nable cuda gpu with cuda path applications cuda 2 2 The other options available can be browsed via the help option Some of them are listed here and they can be of course combined with each other when it does make sense e disable mpi Force not to use MPI during build By default the config ure will try to detect ifthe compiler has some native MPI capabilities If not MPI will be automatically disabled e enable debug Creates a slower version of the executable in which any of the array allocated is filled by NaN after its boundaries Useful to detect runtime errors during developments e with memory limit lt mem gt Creates a version of the executable which abort the run if one of the processes al
14. e tridiagonal base be stored in memory and this limits the number of steps that one can perform The parameter in iat_absorber an integer number is the position starting from one in the posinp xyz file of the absorber atom in L_absorber this parameter is the angular moment of the multipole com ponent of the interaction operator that one wants to consider It is 1 for dipole interaction 2 for quadrupolar in Gabs_coeffs i i 1 2 in L_absorber 1 This are the components of the interaction operator in Cartesian coordinates These correspond to the BigDFT internal representation of tensors which is based on spherical tensor For dipolar interaction these will be therefore the three x y z components In this present version the input are real numbers Entry for the imaginary parts will have to be added in the future versions for magnetic dichroism in potshortcut This parameter determines how the local potential is ob tained If it is set to 1 the potential is obtained from the superposition of atomic charges If it is set to 2 the potential is read from a previous SCF calculation In this case the program searches in the current work directory the file b2B_xanes cube Such file contains the atomic positions and the lo cal potential of a previous SCF run and can be created setting the parameter in output_grid to 2 in the file input dft The SCF potential is interpo lated using interpolant wavelets to get the potential on
15. econd line contains the boundary conditions specified by the keywords free for free boundary conditions periodic for periodic boundary conditions or surface for surface boundary conditions where the x and z direction have periodic boundary conditions and the y direction free boundary conditions The keywords periodic and surface have to be followed by 3 real numbers giving the length of the orthorhombic periodic cell In the case of surface boundary condition the second of these numbers is ignored The following lines contain the name of the chemical element followed by the 3 Cartesian coordinates Chemical elements are identified by their pseudopoten tial If a element is for instance denoted by Si the element will be described by the file psppar Si which has to be present in the working directory of the BigDFT run A silicon atom could however also be denoted by Si_lda if there is file psppar Si_Ida BigDFT supports the GTH and HGH pseudopotentials in the format which can be downloaded from the ABINIT website www abinit org The 3 Cartesian coordinates can be followed by optional additional information In the case of spin polarized calculation the polarization in the region around the atom can be given by an integer In addition the charge in the region around the atom can be specified Finally it can be specified if the atom is fully or partially fixed during a geometry optimization f stands for comp
16. er which serve to determine the radius for the low high resolution sphere around the atom Values are typically of the order of 5 for crmult and of the order of 10 for frmult ixc integer specifying which exchange correlation functional will be used The Abinit conventions are used and detailed information can be found on the Abinit Web page www abinit org documentation helpfiles for v5 8 in put_variables varbas html ixc Here is only a short summary of some wide ly used functionals 1 Pade LDA from Abinit XC library 11 PBE from Abinit XC library 020 Pade LDA from libXC 101130 PBE from libXC 116133 PBEsol from libXC 406 PBEO Hybrid functional local part from libXC teger refers to a functional from ABINIT library and a negative one from libXC library In this case you concatenate the codes for the exchange part and for the correlation part see the table in appendix A 2 A positive in ncharge elecfield Total charge of the system and uniform electric field E E E in units of Ha Bohr A positive ncharge means that electrons are taken away The electric field can not have components in periodic direc tions e g Ex E 0 in case surface BC is used nspin mpol nspin 1 closed shell system without spin polarization nspin 2 spin polarized system nspin 4 non collinear magnetic system gnrm_cv convergence criterion for the wavefunction optimization norm of the gradient Reasonable values are in most cas
17. es between 1 d 4 and 1 d 5 12 e itermax nrepmax Maximum number of gradient evaluations for a single cycle in a wavefunction optimization and maximum number of cycles At the end of each cycle a subspace diagonalization is done which helps in cases where one has near degeneracy between the HOMO and LUMO orbitals 50 and 2 are usually sufficient e ncong idsx ncong gives the number of iterations in the solution of the preconditioning equation For free boundary conditions 5 is a good value whereas for other boundary condition a value from 0 to 2 is in general suf ficient Large values of ncong lead to a smaller number of iterations in the wavefunction optimization and better forces but each iteration is more costly So an optimal compromise value has to be found idsx gives the history length of the DIIS convergence acceleration in the wavefunction op timization 6 is usually a good value for fast convergence In case of convergence problems it can be advantageous to switch off DIIS by setting idsx 0 The memory requirements grow considerably with large values of idsx If memory is the limiting factor one has to choose idsx smaller than the value which gives the fastest convergence The memguess tool can be used to predict the memory requirements for different choices of idsx e dispersion A non zero values activates an empirical add on treatment of dispersion effects The values 1 2 and 3 specify different switching on func ti
18. ess treats one orbital On machines with slower networks each MPI process should have at least 2 to 4 orbitals memguess lt nproc gt option e nproc Number of MPI processes e o The molecule will be rotated such that the size of the workarrays is mini mal e y Generate a file grid xyz containing the coarse and the fine grid points e GPUtest jnrepz Case of a CUDAGPU calculation to test the speed of 3d operators jnrepj is the number of repeats e upgrade Ugrades input files older than 1 2 into actual format e convert jfrom cube etsf to cube etsf z Converts file from to file to using the given formats e atwf jngz Calculates the atomic wavefunctions of the first atom in the gatom basis and write their expression in the gatom wfn dat file ing is the number of gaussians used for the gatom calculation 21 3 2 Single point or geometry optimization bigdft The MPI version is executed on most machines with the mpirun command fol lowed by the name of the executable which is bigdft in our case The treatment of each orbital can be speeded up by using the mixed MPI OpenMP implementa tion where each MPI processes uses several OpenMP threads to do the calculations for its orbitals faster The OpenMP is simply activated by compiling the program with an OpenMP flag and by specifying the number of OpenMP threats by export OMP NUM _THREADS 4 if 4 threads are for instance desired The BigDFT program mo
19. ethod In the PAW method the projector determines the pseudopotential In our case instead the PAW projector is a function of the pseudo potential The implementation of such func tion is detailed here We need to construct the projector for the absorber atom only because the core orbitals don t overlap other atoms For the absorber atom we proceed by calculating the SCF solution in the isolated atom case once using pseudo potentials and another using the all electrons potential Then for these potentials we calculate an almost complete set of eigen functions which will be the basis for the projector We need to do this only for selected angular moments For example for the K edge and dipolar interaction only 1 eigen functions are calculated To calculated the almost complete set of eigen functions we solve the radial Schr dinger equation in the interval 0 R for the lowest N eigenvalues The ra dius R must be large enough to contain the core orbitals We have hard coded it to the value of Sau We have set N 200 With this choice of parameters the maxi mum eigenvalue of the almost complete basis is of the order of XXX Hartree We can compare this value with the maximum eigenvalue representable with wavelets on a grid with spacing d 0 3au for a typical BigDFT calculation which is Emax XXXX Our almost complete basis is indeed complete after this com parison because it covers completely the BigDFT spectra Naming n m and Whim
20. f times To avoid oscillation due to the truncation of the summation a convolution is applied multiplying the u components by the Jackson kernel 3 f b ax 4 3 Use of the code The binary executable for absorption spectra calculation is named abscalc Be sides the usual files which are needed for a basic BigDFT calculation abscalc needs to find in the current work directory at least the file input abscalc This file is specific to absorption calculations and its structure will be explained here The usual files needed for a basic BigDFT calculation and by abscalc are we re call it here input dft posinp xyz and the pseudopotential files psppar XX for all the elements entering in the structure Others files may be needed when one selects the option of importing the self consistent potential from a previous SCF calculation as will be explained later 27 Concerning the posinp xyz file we suggest for a good result to set a large box diameter of about 25 or more and periodic boundary conditions The structure of the file input abscalc is the following The input parameter in iabscalc_type which can be equal to 1 or 2 In the first case the spectra is calculated as a series of Chebychev polynomials while in the second case the Lanczos tridiagonalisation is applied We sug gest the first option because the Lanczos method requires in order to keep orthogonality that all the vectors of th
21. iagonalisation of Hamil tonian ig norbp int Input guess Orbitals per process for iterative diagonalisa tion ig_blocks int int Input guess Block sizes for orthonormalisation ig tol real Input guess Tolerance criterion methortho key Orthogonalisation O Cholesky 1 GS Chol 2 Loewdin rho_commun key Density communication scheme verbosity int Determines the amount of output from little 0 to detailed 2 by default psp onfly T F Switch on the once and for all strategy for calculating the PSP projectors which is faster but more memory demanding considered by memguess The default is on the fly strategy T 20 Chapter 3 Running BigDFT executables 3 1 Estimating the memory usage memguess Before running BigDFT it is recommended to run the memguess program If it runs correctly all input files are available this routine needs a posinp xyz and does not accept a list_posinp file It then allows to estimate the required memory and to find an optimal number of MPI processes for a parallel run For good load balancing each MPI process should roughly treat the same number of orbitals The memguess program prints out the number of orbitals and how many or bitals are treated by each MPI process On a parallel machine with a high perfor mance interconnect network one can choose the number of MPI processes nproc equal to the number of occupied orbitals i e each MPI proc
22. implementation This files contains the following information kptopt This character string specifies the method used to generate the k point mesh auto Automatic generation is used based on the k point density we wish in Fourier space taking into account the symmetries of the sys tem MPgrid A Monkhorst Pack grid using only the special k points tak ing into account the symmetries of the system manual A manual set of k points The following lines depend on the choice of kptopt if kptopt auto One additional line containing a real space length kp trlen BigDFT will automatically generate a large set of possible k point grids and select among this set the grids that give a length of smallest vec tor LARGER than kptrlen and among these grids the one that reduces to the smallest number of k points if kptopt MPgrid Several additional lines The first line should con tains three integers describing the mesh of Monkhorst Pack grid in recipro cal space The second line contains one integer nshiftk that give the number of shift one would like to apply to the MMP grid to obtain the final k point mesh Then the file contains nshiftk lines with three real numbers each giving the shift to apply in reciprocal space if kptopt manual Several additional lines The first contains nkpt an integer giving the number of manually defined k points Then nkpt lines follow with four real values each
23. l 97 version A Filatov amp Thiel 97 version B Perdew Burke amp Ernzerhof exchange solids Hammer Hansen amp Norskov PBE like Wu amp Cohen Modified form of PW91 by Adamo amp Barone Armiento amp Mattsson 05 exchange Madsen PBE like Adamo amp Barone modification to PBE xPBE reparametrization by Xu amp Goddard Becke 86 MGC for 2D systems Bayesian best fit for the enhancement factor JSJR reparametrization by Pedroza Silva amp Capelle Becke 88 in 2D Becke 86 Xalpha beta gamma Perdew Burke amp Ernzerhof exchange in 2D Perdew Burke amp Ernzerhof correlation Lee Yang amp Parr Perdew 86 Perdew Burke amp Ernzerhof correlation SOL Perdew amp Wang 91 Armiento amp Mattsson 05 correlation xPBE reparametrization by Xu amp Goddard Langreth and Mehl correlation JRGX reparametrization by Pedroza Silva amp Capelle Becke 88 reoptimized to be used with vdW functional of Dion et PBE reparametrization for vdW PBE reparametrization for vdW van Leeuwen amp Baerends HCTH functional fitted to 93 molecules HCTH functional fitted to 120 molecules HCTH functional fitted to 147 molecules HCTH functional fitted to 147 molecules Empirical functionals from Adamson Gill and Pople XLYP functional Becke 97 Becke 97 1 Becke 97 2 Grimme functional to be used with C6 vdW term XC_GGA_XC_B97_K XC_GGA_XC_B97 3 XC_GGA_XC_PBE1W XC_GGA_XC_MPWLYP1W XC_GGA_XC_PBELYP1W XC_GGA_XC_SB98_la XC_GGA_XC_S
24. le Whether the stepsize is correct can also be seen from the geopt mon output of the SDCG method In this case the average stepsize in terms of betax should be around 4 after a brief initial period where it is around 8 In contrast to the SDCG method the BFGS method is not 16 very sensitive to the correct stepsize but nevertheless one should try to find reasonable values also in this case ionmov in case of geopt_approach AB6MD the betax line should be replaced by this one It contains an integer value as described in the ABINIT manual Possible values are 6 simple velocity Verlet molecular dynamic 7 quenched molecular dynamic when the scalar product force veloc ity becomes negative the velocity is set to zero The force criterion is tested at each step 8 Nose Hoover thermostat 9 Langevin dynamic adding a friction force and a Gaussian random force on atoms 12 Isokinetic ensemble molecular dynamics The equation of motion of the ions in contact with a thermostat are solved with the algorithm proposed by Zhang J Chem Phys 106 6102 1997 as worked out by Minary et al J Chem Phys 188 2510 2003 The conservation of the kinetic energy is obtained within machine precision at each step 13 Iosthermal isenthalpic ensemble The equation of motion of the ions in contact with a thermostat and a barostat are solved with the algorithm proposed by Martyna Tuckermann Tobias and Klein
25. letely frozen fxz if the atom can only move along the y axis and fy if the atom can only move in the XY plane Below are some examples This hydrogen atom is frozen during the geometry optimization H 1 2 3 4 5 6 This hydrogen atom has a spin up polarization Hy 222 3 456 ST 10 This chlorine atom has an additional electron and no spin polarization Cl 1 2 3 4 5 6 0 1 Same as above but also frozen CE pigor 324 S26 0 sl Jf 2 2 2 The ascii format BigDFT can also use another text file format for input and thus output atomic positions Here is its structure e ist line is arbitrary e 2nd line must contain dxx dyx dyy values e 3rd line must contain dzx dzy dzz values other lines may contain keywords with the syntax keyword followed by a list of keywords separated by commas or blank spaces comments beginning with or x yz name label for atomic position and optional labels dxx dyx dyy dzx dzy dzz values define the box that contains the atoms The for mat allows non orthogonal boxes but BigDFT supports only orthorombic super cells so dyx dzx and dzy must be zeros When the keyword angdeg is used the six values contains the three lengths of basis vectors in dxx dyx and dyy and the three angles bc ac ab in dzx dzy and dzz After the three first mandatory lines all subsequent lines can be comment lines ignored i e empty containing only blanks or beginning with
26. ll the grid points being contained in the union of all these spheres form the low resolution region The high resolution is constructed in a similar way One draws spheres whose radii are the size of the bonding region times the parameter frmult around each atom All the grid points being contained in the union of all these spheres form the high resolution region Default values for the size of the atom and the size of the bonding region are contained in the package The user can however choose different values and these values can be specified by adding an additional line at the end of the pseudopotential parameter files which contain first the alternative values of ermult and then frmult 2 1 3 Visualizing the simulation grid with V_Sim package The grid can be visualized by calling the memguess program with the option y Then the output file grid xyz contains the atomic positions and in addition the grid points which denoted by g if the are in the coarse region and by G if they are in the fine region Whereas the total number of basis functions is nearly independent of the orientation of the molecule the size of some work arrays depends on the orientation When the memguess routine is called with the option 0 it will rotate a molecule that will be calculated with free boundary conditions in such a way that the simulation cell size and hence the size of the work arrays are as small as possible The rotated output is
27. locates more memory than lt mem gt in Gb This version is not necessarily slower that traditional copilation 1 1 2 Make Make the package and create the bigdft executable issuing make The GNU option jn is working with whatever value of n tested up to 16 1 1 3 Install To install the package issue make install It will copy all files to the specified prefix see configure 1 1 4 Clean Clean the source tree of the make action by make clean 1 2 Building a library To avoid to create the binary executable use disable build binary option The main subroutine of the BigDFT package is the call_bigdft routine For a given set of input coordinates and input parameters it returns the total energy and the forces acting on the nuclei The BigDFT f 90 main program calls the call_bigdft routine and can also do a geometry optimization by calling the geopt routine which in turn calls again call_bigdft For other standard applications other main programs exist At present main programs to do a vibrational analysis saddle point search and global optimization have been developed Users are en couraged to write their own main programs for specific applications The BigDFT API will be the object of a forthcoming chapter Chapter 2 File formats and conventions in BigDFT 2 1 The basis set 2 1 1 One dimensional functions The maximally symmetric Daubechies family of degree 16 is used to represent the Kohn Sham orbi
28. m_cv The stop criterion on the residue of potential or density norbsempty Tel The number of additional bands and the electronic tem perature e alphamix alphadiis The multiplying factors for the mixing and for the elec tronic DIIS Each diagonalisation step is done with a direct minimisation scheme at a fixed potential The parameters of this minimisation are the usual ones in the input dft file The overall convergency of the diagonalisation is not very sensitive to the good convergency of the direct minisation steps Thus a itermax value of 5 in input dft is often advicable 2 7 The input file input perf This file is used to specify values in order to optimize the performances of BigDFT If this file does not exist default values are used On the contrary to other BigDFT input file this file has optional key value entries The keywords are not case sensitive Keywords can specify options as e debug T F The debug mode is enable mainly for memory profiling e fftcache ncache size Specify the cache size for FFT in kBytes e accel NO CUDAGPU OCLGPU Specify the use of CUDA resp OCL versions of various subroutines For CUDA a GPU config file is needed 19 blas T F Use or not the CUBlas acceleration projrad real Radius of the projector as a function of the maxrad exctxpar key Exact exchange parallelisation scheme ig diag T F Input guess T Direct F Iterative d
29. mitted At each NEB iteration the driver script is run It creates the directories for forces calculations if needed create the input files using the include script run the jobs grep the energy and the forces and return The script file NEB_include sh must contain the following shell functions init_jobs run once after directory creation It can be used finalise_jobs run once after all jobs have finished wait_jobs run each time the script poll all replicas for completion make_input run once in each scratch directories to create the file pos inp xyz from the NEB restart file and the initial ones run_job run once in each scratch directories to start the force calcula tions In the example it runs the replicas directly on the host machine but in this shell function a submission file may be created instead and submitted to the queue system for later run check_job is used periodically by the driver system to check if a job is finished or not It must return different values 1 if the job is still not started maybe be still in the queue for instance O if the job is running but not finished yet 1 if it exited with success and 2 if it exited with a failure grep_forces is run once after job termination to get the energy and the forces Energy must be in Hartree and forces in Hartree per Bohr For NEB purposes users of BigDFT usually have only to modify the run_job functions from the examples t
30. nitors during a run the memory utilization and the time spent in various subroutines Detailed information is written in the files mal loc pre and time pre At the end the program checks whether the number of deallocations was equal to the number of allocations and whether the total memory went back to zero If this should not be the case please send a bug report to the developers of BigDFT 3 3 Doing a path minimization NEB A NEB implementation is present in the BigDFT package thanks to the initial routines provided by Carlos Sbraccia This implementation is launched with the help of the NEB executable This program is responsible for initialising the path from the two minima and then to run the path minimisation using BigDFT to compute the forces Each replica is launched by an instance of bigdft and not in the same program NEB This allows to run the NEB algorithm on a super computer with a queue system each replica going separately in the queue The process of running the different bigdft instances in different directories and wait for their completions is done by two shell scripts NEB_driver sh and NEB_include sh The first one is very generic and should not be touched by users It is provided in the src directory of the package The second one must be adapted by users to suit to their running machine One example is provided in the tests NEB directory of the package The NEB executable can be run from everywhere but the scri
31. noisy Noise is present because of the underlying integration grid The pa rameter frac_fluct specifies how small the forces should become compared to the noise level to stop the geometry optimization Values in between 1 and 10 is are reasonable for this parameter A value of 2 means that the geometry optimization will stop when the largest atomic force is compara ble to the 2 times the average noise in the forces themselves For values of frac_fluct smaller than 1 one can under certain circumstances obtain better relaxed geometries but one risks that the geometry optimization will not con verge since the forces are too noisy In such a case one should closely mon itor the progress of the geometry optimization by looking at the posout files which are written at each step of the geometry optimization and at the geopt mon file e randdis This parameter allows to add random displacements with ampli tude randdis to the atomic positions in the input file posinp This can for instance be useful to break degeneracies which would lead to convergence problems in the wavefunction optimization in highly symmetric structures e betax This is the stepsize for the geometry optimization This stepsize is system dependent and it has therefore to be determined for each system If the VSSD method is used one can start with a small stepsize of around 1 and VSSD will suggest than a better value for betax in the last line of the geopt mon fi
32. o adapt it to their machine The output files are 23 e job_name NEB dat A three column file containing for each replica a re action coordinate the energy in eV and the maximum value of remaining perpendicular forces This file is updated at each iteration e job_name NEB log A file giving for each NEB iteration the value of the highest replica in eV and the value of the maximum perpendicular forces on all atoms and replicas job_name NEB restart A file with the coordinates of each replicas This file is generated by the NEB executable at each iteration and used by the driver script to generate the posinp files 3 4 Doing frequencies calculations frequencies Using the executable frequencies it is possible to calculate the vibrational prop erties of a system by finite difference The atomic system is moving for each direc tion and each atom in a small step in order to calculate the Hessian matrix by a finite difference scheme A checkpoint restart is implemented The file frequencies res contains the previous calculations 3 4 1 The input freq file There are three parameters e freq_alpha to determine the step for frequency step i e alpha hx al pha hy alpha hz e freq_order which determines the order of the finite difference scheme 1 calculates Fu a N 3 1 calculates Kan h Sm 2 calculates ee this is the default 1 3 calculates Be nz xo h f xo 2h
33. ons using the convention of Q Hill and C K Skylaris Proc R Soc A 465 2009 669 e InputPsild output_wf output_denspot InputPsild specifies how the input guess wavefunction is generated InputPsild 2 Random numbers are used as input guess This is of course a poor input guess which will need many iterations of the wavefunction optimization and might even lead to divergences InputPsild 1 The input wavefunction is imported from the CP2K code which uses Gaussian functions The basis set should be contained in a file named gaubasis dat whereas the coefficients should appear in the gaucoeff dat file Both files are the output files of CP2K code See the H20 CP2K test for an example InputPsild 0 A subspace diagonalization in a minimal atomic basis set is used This input guess should be used in general if one starts a new calculation InputPsild 1 The previously calculated wavefunctions for instance from the previous geometry optimization step are used as input guess Setting InputPsiId to this value does only make sense from within a main program where call_cluster was called previously The old wavefunction is passed to the new call via the data structure restart 13 InputPsild 2 The input wavefunction is read from the wavefunc tion files which contain all the scaling function and wavelet co efficients In case some parameters such as hgridx or crmult have changed compared to
34. orella and Senatore 1D correlation Exchange in 2D Teter 93 parametrization Exchange in 1D Perdew Burke amp Ernzerhof exchange Perdew Burke amp Ernzerhof exchange revised Becke 86 Xalpha beta gamma Becke 86 Xalpha beta gamma reoptimized Becke 86 Xalpha beta gamma with mod grad correction Becke 88 Gill 96 Perdew amp Wang 86 XC_GGA_X_PW91 XC_GGA_X_OPTX XC_GGA_X_DK87 R1 XC_GGA_X_DK87_R2 XC_GGA_X_LG93 XC_GGA_X_FT97_A XC_GGA_X_FT97_B XC_GGA_X_PBE_SOL XC_GGA_X_RPBE XC_GGA_X_WC XC_GGA_X_mPW91 XC_GGA_X_AM05 XC_GGA_X_PBEA XC_GGA_X_MPBE XC_GGA_X_XPBE XC_GGA_X_2D_B86_MGC XC_GGA_X_ BAYESIAN XC_GGA_X_PBE_JSJR XC_GGA_X_2D_B88 XC_GGA_X_2D_B86 XC_GGA_X_2D_PBE XC_GGA_C_PBE XC_GGA_C_LYP XC_GGA_C_P86 XC_GGA_C_PBE_SOL XC_GGA_C_PW91 XC_GGA_C_AM05 XC_GGA_C_XPBE XC_GGA_C_LM XC_GGA_C_PBE_JRGX XC_GGA_X_OPTB88_VDW XC_GGA_X_PBEK1_VDW XC_GGA_X_OPTPBE_VDW XC_GGA_XC_LB XC_GGA_XC_HCTH_93 XC_GGA_XC_HCTH_120 XC_GGA_XC_HCTH_147 XC_GGA_XC_HCTH_407 XC_GGA_XC_EDF1 XC_GGA_XC_XLYP XC_GGA_XC_B97 XC_GGA_XC_B97_1 XC_GGA_XC_B97_2 XC_GGA_XC_B97_D 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 160 161 162 163 164 165 166 167 168 169 170 34 Perdew amp Wang 91 Handy amp Cohen OPTX 01 dePristo amp Kress 87 version R1 dePristo amp Kress 87 version R2 Lacks amp Gordon 93 Filatov amp Thie
35. pts NEB_driver sh and NEB_include sh must be in the same directory It takes arguments from the command line using a redirection from a file is good practice See the test s NEB directory of the package NEB lt input The file input contains a list of variables relevant to the NEB algorithm Some of them are explained here e scratch_dir is where the NEB_driver sh script will create the directories where to run the bigdft instances It is usually on a local disk e job_name a name that will be used to named all generated files and directo ries 22 climbing when set to TRUE the highest replica follow opposite parallel forces in addition to the usual perpendicular ones optimization when set to TRUE tries to optimize the geometry of the first and last replicas if not local minima already minimization_schene specifies the scheme used to minimze the NEB The possible values are steepest_descent fletcher reeves polak ribiere quick min damped verlet and sim annealing tolerance is a criterion not to take into account in the list of moving atoms the ones that move less than this value between the first and the last replica convergence is the stop criterion on perpendicular forces in eV A num_of_images is the number of desired replicas _config the file describing the first and the last replica in BigDFT 1 3 the file extension xyz must be o
36. r 4 2 Introduction The interaction between matter and a photon described by a wave vector k and polarisation is written discarding elastic Thomson scattering the A r term in the Schr dinger equation which becomes dominant off resonance discarding also the spin magnetic field interaction 2 and using CGS units as 2nfic 1 e Hin OV a keexp ik r ax sexp ik r pP 4 1 25 where r is the electron coordinate p the kinetic moment Ars Age are creation annihilation photon operators kc and V is the space volume The photon absorption is a first order perturbative process whose probability per unit of time w for photon with polarization state is obtained from Hin matrix elements by the Fermi golden rule For one photon in the V volume WERS hoVm 2 3 EIERN lt nlexplik r p e 0 gt 8 ia E 42 n where lt n is a complete set of eigenstates of matter lt 0 being the initial state En is the energy difference between lt n and lt 0 and we have retained only the resonating denominator The absorption cross section is the quantity directly ob servable in experiment It is obtained from the above equation dividing by the photon flux c V 25 coho Ayi n exp ik r p e 0 gt 8Gi En 4 3 homc The matrix element can be simplified in terms of r by using the following identity for p exp ik r p Likorto H re 4 4 MUH r e ilH k
37. rr e 2 Such substitution leads to cofo 2r adiof fo 4 5 with fho lt nli gt Sfo E 4 6 where l Fey e LA s 4 7 2 is the state resulting from the application of the interaction operator on the ground state The f i is formally determined by its distribution moments which can be computed by iterative applications of the Hamiltonian f f E E 4dE lt i H i gt 4 8 26 To avoid numerical ill conditioning coming from the numerical linear depen dence of high order powers we proceed as explained in 3 The H Hamiltonian is rescaled and shifted to a new H whose spectra is comprised within 1 1 lt n H n gt 1 1 The overlap integrals between the spectra and Chebychev polynomials 7 x is calculated by recurrence The Chebychev polynomials T x are Ta x cos n arccos x and satisfy the recurrence relation Tm 1 x 2xTin x Tn 1 x The overlap integral are m f FOTO lt ATH gt and can be obtained by applying the recurrence relation to the left side of i gt state Tm 1 H i a 2H T H i gt Tu li gt and then obtaining the scalar product The spectra is finally recovered as uo 25 nTn x anvl x where b and a are the shift and the scaling factor respectively involved in the H H transformation The spectra can be obtained up to an arbitrary degree of resolution by applying iteratively the recurrence relation for a sufficient number o
38. s and machines output_wf 3 The output wavefunctions are written at the end of the wavefunction optimization into ETSF binary files This format is portable between compilers and machines since based on NetCDF Par allel IO are taken into account output_denspot 0 No output density is written output_denspot 1 Output electronic density is written in the pot format of V_Sim into the file electronic_density pot Deprecated use 11 instead output_denspot 2 In addition to the electronic density the poten tial local_potential pot and its components ionic_potential pot and Hartree _potential pot are also output in plain text pot files Depre cated use 12 instead output_denspot 11 Same as output_denspot 1 but files are writ tent in ETSF file format portable binary format based on NetCDF 14 output_denspot 12 Same as output_denspot 2 but files are writ tent in ETSF file format portable binary format based on NetCDF output_denspot 21 Same as output_denspot 1 but files are writ tent in cube file format plain text output_denspot 22 Same as output_denspot 2 but files are writ tent in cube file format plain text e rbuf ncongt Far reaching tails of the wavefunctions decaying into the vac uum are added in a perturbative treatment if the variable rbuf is set to a strictly positive value This allows to do a calc
39. t is advised to create a compi lation directory either inside or outside the source tree Lets call this directory compile gFortran for instance One starts the configure from there source tree path configure One can tune the compilation environment using the following options e FC Specify the compiler including MPI aware wrappers e FCFLAGS Specify the flags like the optimisation flags to pass to the com piler default are g 02 for GNU compilers e Linear algebra options with ext linalg Give the name of the libraries replacing BLAS and LAPACK default none specified Use the 1 before the name s with ext linalg path Give the path of the other linear algebra libraries default L usr 1lib Use the L before the path es e Accelarators enable cuda gpu Compile CUDA support for GPU computing with cuda path Give the path to the NVIDIA CUDA tools de fault is usr local cuda with nvec flags Specify the flags for the NVIDIA CUDA Com piler enable opencl Compile OpenCL support for GPU computing com patible with nable cuda gpu with ocl path Give the path to the OpenCL installation directory default is usr e Optional libraries with etsf io Use ETSF file format binary based on NetCDF for densities potentials and wavefunction files with archives Use compression tar bz2 for position files during geometry optimisation
40. tals The two fundamental functions of this family the scaling function 6 and the wavelet y are shown in Fig 2 1 1 for the one dimensional case To form a basis set these functions have to be centered on the nodes of a regular grid 2 1 2 Wavelet basis sets in three dimensions A 3 dim wavelet basis is made by products of 1 dim functions scaling function and 7 wavelets can be centered on the nodes i j k of a regular 3 dim Cartesian grid They all are products of 1 dim scaling functions and wavelets 15492 x i joz k v2 o i jiwe k We je 2Y 2 px iyo j o z k W je y 2 o x i W y j w z k Vz W x io fo Z h Weia 2 wx do sfw z k Wr Wx idw y o e k Vo y yy ye k LEAST ASYMMETRIC DAUBECHIES 16 scaling function wavelet Figure 2 1 Daubechies Scaling Function and Wavelet of order 16 If a grid point carries the 7 wavelets in addition to the scaling function it belongs to the high resolution region In the low resolution region a grid point carries only a single scaling function In the high resolution region the resolution is doubled in each direction with respect to the low resolution region The grid spacing is specified by the parameters hgridx hgridy hgridz The low resolution region is constructed in the following way Around each atom one draws a sphere whose radii are the size of the atom times the adimensional parameter crmult A
41. the previous run the wavefunctions will be also transformed to the new parameter set Use also this value to restart from ETSF file format wavefunction etsf nc InputPsild 10 The same as 0 but activates the Gaussian help after convergence the wavefunctions are projected onto the localised basis set used for the input guess and Mulliken Charge Population Analysis MCPA is performed on this basis The user has the possibility to perform MCPA with separate basis set functionality to be added InputPsild 11 Restart with Gaussian approximation contained in the restart data structure InputPsild 12 The input wavefunction is read from the wavefunc tions gau file which contains an approximation in a minimal Gaussian basis set of the previously calculated wavefunctions output_wf 0 Do not write the wavefunctions to disk output_wf 1 The output wavefunctions are written at the end of the wavefunction optimization into plain text files If InputPsild was greater than 10 the wavefunction will be written in the Gaussian ap proximation into a single wavefunctions gau file otherwise into wave function files Writing a wavefunction file for each orbital can take a considerable amount of time and disk space output_wf 2 The output wavefunctions are written at the end of the wavefunction optimization into Fortran binary files This format is not portable between compiler
42. the target grid The SCF grid of the previous run must have less points than the target grid This is easily the case because in XANES calculations one usually sets a large diameter for the box The fact of choosing a large box limits the spurious oscillations due to bands effects and to interference with the replica of the absorber atom in the other Bravais cells inYonsteps this is the total number of Hamiltonian applications The larger this number the better the resolution you get in the calculated spectra 4 3 1 Note about the absorber pseudopotential In the actual version of the code 1 4 for the absorber pseudopotential one must use a Z 1 approximation For future versions of the code the development of 28 on the fly pseudopotentials corresponding to the excited electronic structure is foreseen 4 4 Description of the output In case of in iabscalc_type equal to 1 the spectra is written in the current work di rectory with the name cheb_spectra_ NUM where NUM is the number of Cheby chev components The number of components increases at each Hamiltonian ap plication and during the calculation several partial results with increasing number of components are written 4 5 The reversed PAW method The BigDFT code in the actual version as well as in the previous ones uses a norm conserving pseudo potential From this pseudo potential we extract the cor responding PAW projector This is the reverse of the PAW m
43. the three first being the coordinates in reciprocal space in 0 0 5 and the fourth bing the weight If the sum of all weights does not equal to one weight values are renormalised After the regular k point mesh for the self consistent loop one can define in addi tion a specific path of k points to be used to compute a band diagram This requires also to define Davidson usage in the input dft file The band structure is define if the following lines are present bands keyword a character string containing bands the next line contains one integer nseg the number of different segments for the path 18 e the next line contains nseg integer values defining the number different k points to be generated in each segment e then follow nseg 1 lines containing each three floating point values with the coordinates of the vertices in reduced coordinates of the Brillouin zone 2 6 The input file input mix If this file is present the SCF loop is run with a diagonalisation scheme instead of the direct minimisation scheme This file controls how the density or the potential is mixed between each iteration of diagonalisation It must contain the following lines e iscf An integer giving the mixing scheme and the mixing target It follows the ABINIT convention For values lower than 10 the potential is mixed while for values greater than 10 the density is mixed itrpmax Maximum number of diagonalisation iterations e rpnr
44. ulation with some moderate value of crmult and then to extrapolate to the limit of large crmult This procedure is not variational and gives too low energies The true energy is in between the two energies and in general much closer to the extrapolated energy This procedure can also be used to judge whether the chosen value of crmult is large enough for a certain required precision rbuf gives the amount by which the radii for the coarse resolution region are increased in atomic units ncongt gives the number of iterations used in the perturbation calculation Reasonable values for ncongt are around 30 e norbv nvirt nplot Usually unoccupied orbitals are not calculated since they are not needed for the total energy and other physical properties of the electronic ground state Putting norbv to a non zero value will result in the calculation of norbv virtual orbitals in a postprocessing routine after the occupied orbitals have been calculated which uses Davidson iterative treatment nplot of these orbitals will be written in the virtual pot files and nplot if that many exist of the highest occupied orbitals will be writ ten in the orbital pot file The Kohn Sham eigenvalues are written in the ordinary output file nvirt corresponds to the actual number of orbitals the convergence process is taking care of during the Davidson iterations e disable sym a logical to set to T to disable the usage of symmetry op erations in the c
Download Pdf Manuals
Related Search