LMTO MAGNONS - Max-Planck
Contents
1.2. m G 0 00000E 00 00000E 00 00000E 00 Delta Eny Ekap 5000E 01 for Fe 0000000E 00 0000000E 00 for 4s state 0000000E 00 0000000E 00 for 4p state 0000000E 00 0000000E 00 for 3d state Delta Eny Ekap 1 000 for Fe 0000000E 00 0000000E 00 for 4s state 0000000E 00 0000000E 00 for 4p state 0000000E 00 0000000E 00 for 3d state Delta Eny Ekap 2 500 for Fe 28 0000000E 00 0000000E 00 for 4s state 0000000E 00 0000000E 00 for 4p state 0000000E 00 0000000E 00 for 3d state MAGNETIC FIELD POLARIZATION m G 0 00000E 00 00000E 00 00000E 00 Delta Eny Ekap 5000E 01 for Fe 0000000E 00 0000000E 00 for 4s state QO000000E 00 0000000E 00 for 4p state 0000000E 00 0000000E 00 for 3d state Delta Eny Ekap 1 000 for Fe 0000000E 00 0000000E 00 for 4s state 0000000E 00 0000000E 00 for 4p state QO00000E 00 0000000E 00 for 3d state Delta Eny Ekap 2 500 for Fe 0000000E 00 0000000E 00 for 4s state 0000000E 00 0000000E 00 for 4p state 0000000E 00 0000000E 00 for 3d state kkKAK K DELBND finished CPU time 8170 590 FKK kk k K K K K K 2K 7 9 Calculating Susceptibility To find out the dynamical charge and spin susceptibility matrix 4x4 watch out for the following printout xxkkkkk CHIMAT started CPU time 8170 950 Kk kk k k KKK K K COMPLEX SUSCEPTIBILITY Chi q G 0 g G 0 omega Q VECT
3. fe dos idos lt dosfile gt 2 fe scf iscf lt scffile gt 2 out iout lt outfile gt Other Data for Mag Pack 0 9 0 1 nff nef de 16 16 16 32 ni n2 n3 nc 20 20 20 0 02 0 04 0 10 nfft1 nfft2 nfft3 epsr epsg kbz bzr Additional input files O fe hub ihub lt hubfile gt O fe hop ihop lt hopfile gt O fe opt iopt lt optfile gt 2 fe enr ienr lt enrfile gt 2 fe pnt ipnt lt pntfile gt This file is unique and is used to perform self consistent calculations for all q G w u Only a few lines are treated differently from the input to NMTASA PLW programs It will be explained below 4 1 Control Parameters xxx Band Structure Calculation of bcc Fe x xx Control Parameters 3 lift 1 lmto 0 nsph 1 lrwf 0 npfr This set of parameters should be exactly the same as used in the NMTASA PLW calculation Note that the keyword Irwf should be set to 1 when preparing the SCFFILE by NMTASA PLW 4 2 Exchange correlation functional Exchange Correlation Code 14 1 2 3 by VBarth H Gunn L Jan W 10 The exchange correlation should be the same as used in the NUTASA PLW calculation 4 3 Iterative Procedures Limits and Accuracies Iterative Procedure Limits and Accuracy 50 0 2 0 2 1 D 4 6 0 0 3 0 3 niter mix mag eps lbr ibr mixb mixh This set of parameters has the same meaning as in the NMTASA PLW calculation but they are used for calculating changes in cha
4. in eV om_p x 6 599383 om_p y 6 599383 om_p z 6 367504 of fully filled bands 2 input 0 of bands crossing Ef 4 input 9 Energy bands at the Gamma point for spin up states are 18539 63131 63131 63131 71606 71606 3 1184 3 1184 3 1184 3 1976 3 1976 3 8412 3 8412 3 8412 4 6837 5 6510 5 6510 5 6510 7 7848 7 8404 7 8404 12 028 12 028 12 028 12 315 12 315 12 315 Energy bands at the Gamma point for spin dn states are 20076 76970 76970 76970 91505 91505 3 1747 3 1747 3 1938 3 1938 3 1938 3 8547 3 8547 3 8547 4 6752 5 6774 5 6774 5 6774 7 8220 7 8220 7 8418 12 010 12 010 12 010 12 294 12 294 12 294 LR information of fully filled bands nff 0 LR information of bands crossing EF nef 6 Roo BZINT finished CPU time 290 1100 aa kkk kkk This is the end point for module ELECTRONS Structure constants energy bands and wave functions are stored in CONFILE BNDFILE and PSIFILE and are ready for use as shareable files The execution tranfers to the module MAGNONS see file magnons f 7 6 Preparing Integration Weights The first step performed by the module MAGNONS is the calculation of the k space integration weights necessary for calculating induced charge density and magnetization The information about symmetry of the q vector is printed out It includes number of equivalent operations group for this vector as well as its symmetry star x xkkkkk CHITET started CPU time 290 1100 FKK k k K K
5. K 2K K 2K Symmetry analysis for q 0000E 00 0000E 00 7500 1 group of q vector contains 8 elements 2 star of q vector contains 6 elements Calculated average square of electron velocities lt Vx 2 gt 7472877 lt Vy 2 gt 7472877 lt Vz 2 gt 6956963 26 Calculated bare plasma frequencies in units eV om_p x 6 599383 om_p y 6 599383 om_p z 6 367504 UP states TEST TOSK DOSK 1 000000 2 551241 5 341420 UP states TEST TOSQ DOSQ 1 000000 2 551241 5 341420 DN states TEST TOSK DOSK 1 000000 1 448759 1 707500 DN states TEST TOSQ DOSQ 1 000000 1 448759 1 707500 AL states TEST TOSK DOSK 2 000000 4 000000 7 048919 AL states TEST TOSQ DOSQ 2 000000 4 000000 7 048919 When the integration weights are calculated the following test lines are printed out UP states TEST TOSK DOSK 1 000000 2 551241 5 341420 UP states TEST TOSQ DOSQ 1 000000 2 551241 5 341420 DN states TEST TOSK DOSK 1 000000 1 448759 1 707500 DN states TEST TOSQ DOSQ 1 000000 1 448759 1 707500 AL states TEST TOSK DOSK 2 000000 4 000000 7 048919 AL states TEST TOSQ DOSQ 2 000000 4 000000 7 048919 Susceptibility calculation states UP UP ibnd ibnd1 chi V chi F chi1 d chi2 d 1 1 0000000E 00 0000000E 00 0000000E 00 0000000E 00 1 2 0000000E 00 0000000E 00 0000000E 00 0000000E 00 1 3 9930953E 03 0000000E 00 9516821E 03 9002467E 03 1 4 2953428E 02 0000000E 00 2827720E 02 2673846E 02 1 5 3
6. create a subdirectory in mag dat For example to calculate x of Fe create magplw dat fe In this subdirectory create the following subdirectories this is default list of the subdirectories listed in the linear response control file LRTFILE See also chapter LINEAR RESPONSE CONTROL FILE LRTFILE e magplw dat fe INP subdirectory for storing input files like main input control file INIFILE structure control file STRFILE self consistent charge density file SCFFILE linear response control file LRTFILE etc e magplw dat fe DRO subdirectory for storing the files with changes in charge density and pseudodensity All names for these files will be given automatically e magplw dat fe WGT subdirectory for storing the files containing the k space integration weights All names for these files will be given automatically Note that the directory INP contains the files which are unique for all q G points and frequencies w All other directories contain the files which are specific for any particular q G point w and field polarization Since there are too many files 4 x Nyya x Nu their filenames are constructed automatically For example the rule to construct the name for DROFILE is i three first letters are dro ii q point number iii G vector number iv w v polarization Polarization w 1 abbreviated as m u 0 abbreviated as z and u 1 abbreviated as p If scalar potential 9999 field is used for ch
7. from the INI STRFILEs lt lt lt INPUT INIFILE READ gt gt gt xxx Band Structure Calculation of bcc Fe lt CONTROL PARAMETERS gt lmto 1 Unscreened LMTO is on lrwf 1 Adjust Phi to Veff only npfr 0 Atomic forces are off lt EXCHANGE CORRELATION gt ixc 14 Vosko Wilk Nussair GGA91 lt ATOMIC DATA gt natom 1 of atoms in unit cell nsort 1 of atoms of different type par 5 425000 lattice parameter in a u nspin 2 including spin polarization norbs 1 without spin orbit coupling nkap 3 of kappas in valence panel Etaili 50000E 01 Hankel tail energy Etail2 1 0000 Hankel tail energy Etail3 2 5000 Hankel tail energy lt OUTPUT DATA gt icon 2 read str const from INP fe con iftr 0 no storage fourier con ibnd 2 save tetra bands in INP fe bnd idos 0 no d o s calculation ipot 0 no storage of full potential iscf 1 save full density in INP fe scf iout gt 1 print current output lt OTHER DATA gt nff 0 of bands below EF nef 9 of bands crossing EF ndiv 16 16 16 tetr mesh for valence panel ndic 32 32 32 tetr mesh for semicore panels 21 lt ADDITIONAL INPUT FILES gt ichub 0 ichop icopt icenr gt Oo ooo icpnt gt lt lt lt INPUT LRTFILE READ with with no Hubbard correction no hopping matrix int after INP fe enr with no optical properties with dense grid for
8. is not then the iterational cycle will continue You can also use this when it is necessary to improve convergency of the induced charge density Suppose it is self consistent with the accuracy say 107 In order to improve the convergency reset eps from 107 to say 107 in the INIFILE The iterational cycle for given displacement and polarization will be renewed e lbr switches on the Broyden mixing If 1 then Broyden is OFF if 0 then Broyden is ON for 1 0 component of dp r if 1 2 then Broyden is on for l le lbr component of Drho r It is highly recommended to set lbr to the Imax value of spherical harmonics expansions for the charge density and potential It is controlled by the parameter LmaxV see below and it is usually 4 for ASA or 6 for PLW Broyden restarts every time after NTERMAX iterations Parameter NTERMAX is in PARAM DAT Usually NTERMAX 15 e ibr starts Broyden after I ibr iterations If br 0 then start immediately If br gt 0 then first I ibr iterations will be done with the linear mixing scheme where the mixing parameters miz and mag are specified above e mixb this is initial guess for Jacobian which is closely related to mixing parameter migs in the linear mixing scheme It was found that mixb cannot be small and it is usually of the order 0 3 0 4 11 e mizh this is linear mixing parameter for higher components gt lbr of the charge density Since it is assumed that these components do not influence
9. is printed Also the number of k points controlled by parameters nw1 nw2 nw3 which will be used to find the integration weights in the linear esponse calculation is printed xkkkkk STRMSH started CPU time 3 290000 ER OR A OR KK k 48 elements discovered 145 k points generated 897 k points to weight Position 00000E 00 00000E 00 00000E 00 for Fe LMTO basis set is expanded in spherical harmonics up to Lmax 6 Charge density is expanded in spherical harmonics up to Lmax 6 Non zero elements allowed by symmetry are the following l 0 m 0 l 4 m 4 0 4 l 6 m 4 0 4 23 Total of non zero components found 7 l Nplw Ecut Ry RH S H S GH S RH S A 2346 145 1 00000 1 00000 i eae ta 2346 145 1 00000 1 0000 EEDA 2346 145 99999 1 0000 die ee 2346 145 99997 99997 22 Ar 2346 145 99984 99972 fF B 2346 145 99944 1 0001 ASS 2346 145 99833 1 0030 Result from VECGEN for direct reciprocal spaces gt Rmax 2 616219 Accuracy 2938758E 23 of vectors 169 Gmax 6 592308 Accuracy 1165861E 16 of vectors 603 Smax 4 473601 Accuracy 2950956E 23 of vectors 749 Min energy for using Evald s method 4 983823 Ry Total of connecting vectors found 1 Minimum difference between k G 2 and kappal 2 is 3727432E 01 Minimum difference between k G 2 and kappa2 2 is 7454864 Minimum difference between k G 2 and kappa3 2 is 1 863716 xkkxkkk STRMSH finished
10. much the self consistence loop they are mixed within linear mixing scheme and do not stored for all previous iterations 4 4 Atomic Data Atomic Data They 2d natom nsort nspin norbs 5 425 1 0 lattice parameter v v0 1 is iatom 3 0 05 0 0 1 0 0 0 2 5 0 0 Nkap Ekap ikap This set of parameters should be exactly the same as used in the NMTASA PLW calculations Note that an option switching spin orbit coupling is not yet available for the linear response magnon program Note also that number of spins must ALWAYS be set to 2 when using MAGASA PLW for Fe SBRBESRTTRSSS RSS AS ARAS 26 D0 18 D0 1 1 0 0 5D0 55 847 z zcor lr icor ispl split mass 2 349 1 D0 1 D0 0 DO mt sphere rou sphere weight rloc 626 lmax t l1max b lmax v This set of parameters should be the same as used in the NMTASA PLW calculation e for Nb title for every atom Note that this character string maximum 10 letters will be read and widely used in the output file Therefore it is recommended to use the format for EL e z atomic number e zcor deep core charge e lr 0 for non relativistic calculations 1 for scalar relativistic valence states e icor this parameter has no effect for linear response calculation e ispl split both these parameters have no effect for linear response calculation e mass atomic mass of the element as in the periodic table e mt sphere non touching muffin tin sphere
11. out icon lt confile gt itmp lt tmpfile gt ipsi lt psifile gt itmp lt tmpfile gt ibnd lt bndfile gt ipot lt potfile gt iptn lt ptnfile gt idos lt dosfile gt iscf lt scffile gt iout lt outfile gt This section gives the names of the files which are used by the program The consequence of the files is the same as in the INTFILE of the NMT programs The meaning of the control parameters is the following 13 0 file is not created or created as temporary 1 file will be opened as a new one and saved if file exists the execution will be terminated 2 file already exists and will be read if file does not exist the switch will be automatically set to 1 6 the contents of the file will be printed out to the terminal channel 6 icon lt confile gt this is the structure constants file which will be created by MAGASA PLW at the beginning Note that the CONFILE created by MAGASA PLW cannot be used by the program MAGASA PLW Keep structure constants in the INP directory to make them shareable when running different q points spontaneously Always set icon 2 itmp lt tmpfile gt this parameter is not used by MAGASA PLW ipsi lt psifile gt this is the file which contains wave functions on the grid of k points in the IBZ Keep wave functions in the INP directory to make them shareable when running different q points spontaneously Always set ipsi 2 iscr lt scrfile gt this is the first character stri
12. package One more parameter statement is added in PARAM DAT file of the MAGASA PLW packages com pared to that file of the NMT packages PARAMETER NBNDMAX 15 MAX NUMBER OF BANDS CROSSING EF This parameter sets maximal number of bands allowed to cross the Fermi level The input pa rameter nef cannot exceed this number Do not set NBNDMAX to the large number say maximum possible number of bands is NDIMMAX since it affects the core memory 8 2 Estimation of the needed core memory The storage of the core memory is taken by many arrays For the configuration with 5 6 atoms per unit cell the needed core memory can be of the order 100 Mbyte A very useful option is to link the programs and to get a map file Under UNIX it is xlf o o main exe bloadmap map At the end of the map file there is a total amount of core memory required by the program Disc space significant disc storage is taken by the induced charge density files located in the DRO directory For dealing with the systems of 5 6 atoms per cell a free disc space of the order several Gbytes is desired 30 9 ERROR MESSAGES The description of the error messages in the program is supposed to be essentially extended in the future Below just a few useful hints is given Generally two kind of errors exist in the program warning messages when the program does not terminate and the error messages when the program terminates Normally warning messages mean that the program c
13. with perturbation q s gt gt gt after INP fe pnt SET UP DATA FOR LINEAR RESPONSE CALCULATIONS lt GENERAL SETTINGS gt lift 1p1z 1stn 1d1f mode lt DEFAUL iwgt iphn idsv RHRHROOMH FILE SETTINGS gt idro idro NOrPrRPOOF iout lt lt lt INPUT STRFILE READ Self consistency is Polarizability is Stoner matrix is Changes in phi are on off off off Restart calcuation is off save tetra no storage no storage update ind density update psd density save current output weights gt gt gt Roo Structure Data for bcc Fe x lt CONTROL PARAMETERS gt natom 1 b a c a ibas 0000 0000 in WGT wgt11w500 of phonon modes of ind potential in DRO dro11g0w500m DR0 dps11g0w500m out11w500m in in of atoms in unit cell orthorombicity orthorombicity along b along c ibz icalc istrn ndivi ndiv2 nvec alpha 1 1 1 1 1 1 4 1 300 1 0000 lt PRIMITIVE TRANSLATIONS gt 50000 50000 50000 50000 50000 50000 basis in Cartesian sys automatic BZ choice using built in calc distort cutoff sphere polyhedron generation polyhedron surface grid vectors in Evald method splitting factor there 50000 Rix Rly Riz 50000 R2x R2y R2z 50000 R3x R3y R3z 22 lt BASIS ATOMS IN CELL gt 00000E 00 O0000E 00 00000E 00 lt STRAIN MATRIX gt 1 000
14. 0 00000E 00 00000E 00 00000E 00 1 0000 00000E 00 00000E 00 00000E 00 1 0000 lt INVERSE STRAIN MATRIX gt 1 0000 00000E 00 00000E 00 00000E 00 1 0000 O00000E 00 00000E 00 O0000E 00 1 0000 lt POINT GROUP DESCRIPTION gt ikov C Cubic system lt RECIPROCAL LATTICE gt 00000E 00 1 0000 1 0000 1 0000 00000E 00 1 0000 1 0000 1 0000 00000E 00 lt BRILLOUIN ZONE gt 00000E 00 1 0000 1 0000 1 0000 00000E 00 1 0000 1 0000 1 0000 00000E 00 Cell Volume 79 83057 for Fe Sxx Sxy Sxz Syx Syy syz zx Szy 5zz Rxx Rxy Rxz Ryx Ryy Ryz Rzx Rzy Rzz G1x Gly Glz G2x G2y G2z G3x G3y G3z K1x Kly Klz K2x K2y K2z K3x K3y K3z After executing the INIT subrouitine the execution transfers to the package of programs controlled by module ELECTRONS see source file electrons f This module prepares structure constants energy bands and wave functions necessary for linear esponse calculations 7 2 Preparing Structure Constants STRMSH see source file strmsh f prepares data depending on the crystalline structure Next line gives an information about number of the point group elements found for the lattice The number of k points generated for the main valence panel controlled by parameters nk1 nk2 nk3 the number of k points for all semicore panels is the same as for the main valence panel
15. 000000E 00 Z magnet field PBmO S 0000000E 00 0000000E 00 M magnet filed PBm_ S 1049874E 01 0000000E 00 P magnet field PBm S 0000000E 00 0000000E 00 Induced density and magnetization gt Induced density Drho S 0000000E 00 0000000E 00 Pseudo density Prho S 0000000E 00 0000000E 00 Z Magnetization DMg0 S 0000000E 00 0000000E 00 M Magnetization DMg_ S Q000000E 00 0000000E 00 P Magnetization DMg S 0000000E 00 0000000E 00 Z Pseudomagnetz PMg0 S 0000000E 00 0000000E 00 M Pseudomagentz PMg_ S 0000000E 00 0000000E 00 P Pseudomagentz PMg S 0000000E 00 0000000E 00 Induced charges and magnetic moments gt Induced charge in S_mt 0000000E 00 0000000E 00 Pseudo charge in S_mt 0000000E 00 0000000E 00 Z Magnetic moment Dmom0 QQ00000E 00 0000000E 00 M Magnetic moment Dmom_ 0000000E 00 0000000E 00 P Magnetic moment Dmom 0000000E 00 0000000E 00 xk DELPOT finished CPU time 7418 510 eK k k 7 8 Calculating Induced Charge Density After constructing induced potential induced charge density is calculated This part is controlled by the module DELBND source file delbnd f Changes in Eny which are induced by the perturbation are printed out CPU time 7418 510 Calculation of Linear Response kkKKKK DELBND started FO kkk kk 3kappa Panel 1 MAGNETIC FIELD POLARIZATION
16. 706024 0000000E 00 3389146 3126199 1 6 1 671572 0000000E 00 1 571990 1 466132 1 7 9241452 0000000E 00 9152159 8849650 truncated After the line DELSTR finished is printed out the integration weights are ready They are stored in WGTFILE placed in the WGT directory Note that WGTFILE has a dependence on the wave vector q but not on the displacement and polarization 7 7 Calculating Induced Potential After calculating integration weights and polarizabilities the self consistent cycle begins As a first step here the potential induced by particular displacements and polarizations is calculated This is controlled by module DELPOT see source file delpot f The values of the induced potential induced pseudopotential at the sphere boundary as well as the values of the induced density and the induced pseudodensity at the sphere boundary are printed out ITERATION 1 FOR Q VECTOR O000E 00 0000E 00 7500 kkKAKK DELPOT started CPU time 7246 610 akk k k K K K K K 2K MAGNETIC FIELD POLARIZATION m G 0 00000E 00 00000E 00 00000E 00 27 Data variation atom position 1 Screened potential and magnetic fields for Fe Total potential Dpot S 0000000E 00 0000000E 00 Pseudopotential Ppot S 0000000E 00 0000000E 00 Z magnet field DBm0 S 0000000E 00 0000000E 00 M magnet filed DBm_ S 1449128E 02 0000000E 00 P magnet field DBm S 0000000E 00 0
17. CPU time 5 330000 FIR kk 7 3 Finding Full Potential The full potential is calculated in the same way as it is done in the package NMTPLW As a result the following table is produced xxkkkkk VFULL started CPU time 5 610000 FKK k k K K K K K 2K Input data for Fe in the position T aA gt V up S 1249418 RO up S 2056066E 01 P up S 1063844 PD up S 2023072E 01 P up 0 13 07857 PD up 0 1934313E 02 V dn S 1189130 RO dn S 2110278E 01 P dn S 1553160 PD dn S 2074349E 01 P dn 0 12 83667 PD dn 0 1148676E 02 M S 2 257298 PM S 1 115492 Average potential over the sphere boundaries is 1219274 Average potential in the interstitial region is 1917914E 01 Total charge in the interstitial region must be 9574413 Total charge found via fourier transform is 9574413 Auxilary density renormalization coefficient is 9999995 Magnetization in the interstitial region is 5352703E 01 Total magnetization found in elementary cell is 2 203771 x xkxxkkk VFULL finished CPU time 60 10000 FE CK ACK 7 4 Calculating Energy Bands After constructing the full potential the execution of the ELECTRONS goes to the package of program for solving the eigenvalue problem of the LMTO method It is controlled ba the program BANDS 24 see source file bands f Information about choice of Eny is printed below kkKKKKA BANDS started CPU time 60 10000 FKK k k K K K K K 2K 3kappa spin up panel 1 Band Structu
18. LE PARAM DAT 8 1 Differences with the NMT package 2 aa ee 8 2 Estimation of the needed core Memory ERROR MESSAGES 9 1 Errors connected with PARAM DAT 0 0 00000002 2 eee 9 2 Errors connected with input 2 2 2 e 10 USING MAGLIB LIBRARY 10 1 Program QPNT 11 Acknowledgments 12 COPYRIGHT 29 29 29 30 30 30 31 31 32 32 1 INTRODUCTION The linear esponse linear muffin tin orbital LR LMTO programs described here are designed to perform linear response calculations of the dynamical spin and charge susceptibilities for arbitrary wave vectors q G and frequency w within the methods of time dependent density functional theory TD DFT Refs 1 2 3 4 The development described here is analogous to the linear response calculation of the lattice dynamics and electron phonon interactions which has been described in several publications 5 6 7 8 A short description of the method for calculating dynamical susceptibilities is appeared recently 9 The purpose of the linear response method is to find change in charge density and magnetiza tion induced by a perturbation of a ceratin wave vector q If one is interesting by the dynamical spin or charge susceptibility x r r w the perturbation is given by either scalar external potential Vez r t vexpli q G r wt or external magnetic field Bezt r t dbexpli q G r wt Fixing wave vector q selecting reciprocal la
19. LE but in the root directory 14 4 6 Other Data for MAG pack Other Data for Mag Pack 0 9 0 1 nff nef de 16 16 16 32 ni n2 n3 nc 20 20 20 0 02 0 04 0 10 nfft1 nfft2 nfft3 epsr epsg kbz bzr This set of parameters is similar to what is used in the NMTPLW calculation However there are differences The exact meaning is given below Additional input files 0 0 0 2 2 nff nef number of filled bands in the main valence panel above the semicore and number of bands crossing the Fermi level These parameters must be chosen with more care in contrast to NMTPLW see subsection Notes to k space integration for the detailed description pole parameter connected with the Lorentizan broadening of the Fermi surface integrals The parameter is not used by NMTPLW See also subsection Notes to k space integration nk1 nk2 nk3 divisions of the Brillouin zone along three directions for the tetrahedron integra tion see subsection Notes to k space integration nwl divisions of the Brillouin zone along three directions for finding the weights in the tetra hedron integration Only the first division must be specified two others nw2 nw3 will be calculated according to nk1 nk2 nk3 Note that this parameter substitutes the parameter nc of the NMTPLW See subsection Notes to k space integration for the detailed description ml m2 m3 divisions of the unit cell for the fast Fourier transform Thes
20. LINEAR RESPONSE PROGRAM PACKAGE LMTO MAGNONS USER s MANUAL S Yu SAVRASOV Max Planck Institute fuer Festkoerperforschung D 70569 Stuttgart Germany Department of Physics and Astronomy Rutgers University Piscataway NJ 08854 October 13 2000 Contents 1 2 3 INTRODUCTION INSTALLATION RUNNING MAGASA MAGPLW PROGRAMS 3 1 Configuring MAGASA MAGPLW ti A as hea a Ah ia 32 Terminal input a A REM eh Ge Ree eee Pes MAIN CONTROL FILE INIFILE 4 1 Control Parameters oaoa aa 4 2 Exchange correlation functional aoaaa ee 4 3 Iterative Procedures Limits and Accuracies 0000000004 A At Atomic a A Myke Pa as pa pre bla et iy te Be a 4 5 Input Control Piles ni a nodes Be ea eee ae Ela 4 6 Other Data for MAGe pack ee 4 7 Notes to k space integration ee LINEAR RESPONSE CONTROL FILE LRTFILE EXECUTING MAGASA PLW IN PARALLEL REGIME OUTPUT MESSAGE FILE OUTFILE fly Reading Input Data o 2 6 wipe cp oh eh dowel oe Slee a eke aw Ab a ae 7 2 Preparing Structure Constants o ooa aa ee 7 3 Finding Full Potential a 7 4 Calculating Energy Bands s a esa oaaao 0 02 ee ee 7 5 Integrating over Brillouin Zone 2 0 ee 7 6 Preparing Integration Weights 2 0 e 7 7 Calculating Induced Potential 2 0 20 20 2020000000000 ee 7 8 Calculating Induced Charge Density 0 0 0 2000000202 eae 7 9 Calculating Susceptibility 2 0 0 0 0 a PARAMETER FI
21. MAGASA uses atomic sphere approximation and does not treat interstitial region correctly MAGPLW uses plane wave representation for all the relevant quantities in the interstitial region and therefore much more accurate Note that MAGPLW code is much slower than MAGASA program The purpose of these MAG programs is to calculate x r q G w The basic input to the program is given by the charge density distribution of the unperturbed crystal It is therefore necessary to install and learn how to use the programs NMTASA and or NMTPLW for the original self consistent band structure calculations with the LMTO method This program uses the same plane wave representation for the charge densities and the potentials in the interstitial region The package of programs NMT including three versions NMTASA NMTCEL and NMTPLW is distributed separately and can be downloaded from http www mpi stuttgart mpg de docs ANDERSEN SAVRASOV frames htm Finding x r q G w is divided in two parts which run independently e Preparational part A self consistent band structure calculation with the NMTASA or NMTPLW code must be performed for the original crystal A self consistent charge density file SCFFILE must be created e Self consistency part Self consistent linear response calculation is performed for each wave vector q G frequency w and for either external scalar potential field for finding charge sus ceptibility or for external magnetic field for findi
22. OR 0000E 00 0000E 00 7500 G 0 00000E 00 00000E 00 00000E 00 G O0 00000E 00 00000E 00 00000E 00 FREQUENCY 3674903E 01 Ry 500 0000 meV m Zz p v m 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 Zz 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 p 1 688 1524 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 v 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 0000E 00 xxxkxkkk CHIMAT finished CPU time 8171 000 Kk kk k Kk KK Units used for the susceptibility are the following longitudinal response functions are in st Ry cell For example the non interacting at first iteration longitudinal susceptibility is exactly N Ep Transverse susceptibility is given in the units of magnetic moment For q w 0 non interacting at first iteration x 0 0 should be equal to magnetic moment in the unit cell 29 0000 0000 0000 0000 8 PARAMETER FILE PARAM DAT The dimensions inside the program are defined using PARAMETER statements All parameter stat ments are collected in one file called PARAM DAT and are included in every program using INCLUDE statement PARAM DAT used by MAGASA PLW program is very similar to the file PARAM DAT used by NMTASA PLW programs We therefore refer to the manual of the NMT packages for complete description Here only two differences between the two files will be pointed out 8 1 Differences with the NMT
23. Section GENERAL SETTINGS describes linear response control parameters GENERAL SETTINGS Magnons Response scheme Phonons Magnons gt Plz off Polarizabilty on off gt Stn off Stoner renormalization on off gt Del off 4 Changes in radial functions on off Dyn none Dynamical matrix scheme none hf okal oka e Response scheme This string must be set to Magnons Other option Phonons is used by the program PHNPLW e Polarizability Plz off key must be used e Stoner renormalization is not used by the MAGASA PLW program 18 e Changes in radial wave functions must be switched OFF e Dynamical matrix scheme must be set to none The following section is called default file settings DEFAULT FILE SETTINGS 0 NNOONNOOOMN gt DSF dsf idsf lt dsffile gt WGT wgt iwgt lt wgtfile gt gt TMP tmp itmp lt tmpfile gt gt STN stn istn lt stnfile gt PHN phn iphn lt stnfile gt gt PLZ plz gt iplz lt plzfile gt gt PLZ pls ipls lt plsfile gt gt POT dsv idsv lt dsvfile gt gt POT dpv idpv lt dpvfile gt gt DRO dro idro lt drofile gt DRO dps idps lt dpsfile gt Since there is a lot of different files which are stored for every q G w y it is useful to sort them in different subdirectories Generally there several subdirectories e WGT contains weights for the k space integration The file names will be constructed automat ical
24. an either correct the problem itself or the problem is not important for the execution like an advise to switch on scalar relativism when atomic charge exceeds 21 The error messages always mean that the program can give a wrong result if the input files will not be corrected 9 1 Errors connected with PARAM DAT In all cases the program will terminate if any of the actual parameters exceeds the parameter in the PARAM DAT Then standard error message occurs which states which of the parameters must be increased 9 2 Errors connected with input Some input data can be easily checked like the number of atoms which is read from different input files If there is a mismatch in the input a corresponding message is printed and execution is terminated 31 10 USING MAGLIB LIBRARY 10 1 Program QPNT This program is used to generate irreducible points in the Brillouin zone No special configuring of the program is required Just compile contents of the directory maglib qpnt link object files and get executable make exe The input files to the program are INIFILE and STRFILE The input line also contains three divisions ng1 nq2 nq3 necessary to construct the grid of wave vectors q The output is the number of irreducible q points generated according to the point group and the list of points which can be stored in the output file called PNTFILE An additional option is provided to generate the irreducible points with its minimal lenght This
25. arge susceptibility it is abbreviated as v For example when one finds self consistent response for the q point number 12 listed in PNTFILE G vector 3 w 30 meV and polarization z along magnetization axis the change in charge density DROFILE will be stored in the file drol2g03w30z and will be automatically placed into the directory DRO Analogously names are constructed for DPSFILEs etc 3 2 Terminal input MAGASA MAGPLW programs have several input lines IFOLDER INIFILE SCFFILE STRFILE LRTFILE Q POINT G SHELL FREQUEN RUNMODE RUNTASK The answers must be either given from the terminal or placed into a job file General description of the input lines is given below IFOLDER gives the subdirectory name where all input files are stored For example if the input subdirectory is called INP the string INP must be given to the answer IFOLDER INIFILE is a name of the main control data file The structure of the file is identical to the INIFILE of the NMT package A description of this file and its difference to that of the NMT package will be given in section MAIN CONTROL FILE INIFILE INIFILE must be stored in the subdirectory specified by the parameter IFOLDER SCFFILE is a name of the self consistent charge density file which is calculated by the NM TASA PLW packages SCFFILE must be stored in the subdirectory specified by the parameter IFOLDER STRFILE is a name of the structure d
26. ata file It is the same as used by the NMT package therefore it is not described in this manual STRFILE must be stored in the subdirectory specified by the parameter IFOLDER LRTFILE is a name of the linear response control file This file is specific for linear response calculation and does not contain any information about the compound Its detailed description will be given below in the section LINEAR RESPONSE CONTROL FILE LRTFILE LRTFILE must be stored in the subdirectory specified by the parameter IFOLDER Q POINT is a q point number A list of q points must be prepared and stored in the PNTFILE Use program QPNT located in maglib qpnt for this purpose See section USING MAGLIB LIBRARY for the detailed description Lines in the PNTFILE numerate q points A path to the PNTFILE is contained in the INIFILE Therefore setting Q POINT number from the terminal will result in reading the corresponding line from the PNTFILE Note that Q POINT number must be set as a character string The reason is that all output files containing change in charge densities potentials etc must be named differently for different q points The character string specifying q point will be added to any output filename For example Q POINT 01 means that the first point from PNTFILE will be treated as a perturbation wave vector and the string 01 will be added to any of the output files G SHELL shows which of the reciprocal lattice vectors will be added to q
27. ate q points Therefore setting Q POINT number from the terminal will result in reading the corresponding line from the PNTFILE For convenience place it into the INP directory Always set ipnt 2 e ienr lt enrfile gt this is the file which contains energy bands on the dense grid of k points in the IBZ see subsection Notes to k space integration Keep energy bands in the INP directory to make them shareable when running different q points spontaneously Always set ienr 2 4 7 Notes to k space integration This section explains how to handle with the parameters responsible for the k space integration which is performed to find change in charge density for every vector q This set of parameters involves nff nef number of bands below and crossing the Fermi level pole Lorentzian broadening nk1 nk2 nk3 BZ divisions for integration nw1 nw2 nw8 BZ divisions to find the integration weights and lt enrfile gt file which contains energy bands generated at the grid set up by nw1 nw2 nw3 For semiconductors when there is an energy gap between occupied and unoccupied states the k space integration using misweight free tetrahedron method is equivalent to the special point method In this case set nff to the actual number of bands below the Fermi energy nef to zero pole to zero specify the grid parameters nw1 nw2 nwW3 equal to the BZ divisions nk1 nk2 nk3 ENRFILE which contains the bands generated at the grid nw1 nw2 nw3 will be th
28. atements are contained in the file PARAM DAT See Section PARAMETER FILE PARAM DAT for the detailed description The file PARAM DAT is included into every source code during the compilation time by INCLUDE statement The second comment concerns a scratch file storage To minimize core memory some data during the run are temporary stored in the scratch files To be able to do this a scratch directory must exist on any particular node where execution of the program is performed To create executable file of the MAG program go to the directory mag run 1 Edit PARAM DAT and install the necessary size of arrays Sample PARAM DAT file is provided 2 Edit the file setup f and specify the path to the scratch directory Also check that other set up data match with your local operational system 3 Edit the file timel f and specify the call to the system subroutine to learn CPU time 4 Compile all programs link them to get executable file main exe Under UNIX using AIX XL Fortran Compiler this looks like x1f cOw f to compile only with optimization and suppress ing all warning messages The command xlf cCg f will compile only suppress optimization and provide debugging information To link use the command xlf o o main exe To create a load map use the command xlf o o main exe bloadmap map At the end of the map file a total amount of the core memory allocated by the program is printed out To run MAG for a particular compound
29. d An example of the compound which will be considered below is ferromagnetic iron The input output data files are contained in magplw dat fe directory There exists a library directory maglib It containes a number of auxiliary programs The programs understand input output data files of the MAG codes The contents of maglib is e maglib gpnt program which generates irreducible set of q points according to the input divisions and stores them into the PNTFILE The latter is one of the input files to the MAG codes Note that the perturbation wave vectors q must be chosen to have minimal length one of the options while running QPNT in order to have well convergent plane wave expansions see below All programs and data files are tared and gzipped into 2 files named as magasa tar gz mag plw tar gz and maglib tar gz 1 gunzip magasa tar gz 2 tar x f magasa tar Repeat these steps for magplw dec and maglib dec 3 RUNNING MAGASA MAGPLW PROGRAMS In this section general hints how to run main linear response codes MAGASA MAGPLW are de scribed Section USING MAGLIB LIBRARY will describe how to run auxiliary library routines 3 1 Configuring MAGASA MAGPLW To be able to run MAGASA MAGPLW it is necessary to compile the source data files A few comments must be said here First the maximum size of every array such as maximum number of atoms lmaz etc in the program is declared using the Fortran PARAMETER statement These st
30. e parameters should be the same as used by the NMTPLW calculation epsR epsG accuracy of matching the spherical Hankel functions in real and reciprocal space These parameters should be the same as used by the NMTPLW calculation keyt bzr to accelerate Fourier transforms when calculating interstitial potential matrix el ements set keyt 1 In this case the radius of the cutoff sphere in reciprocal space is set by parameter bzr times the radius of the sphere circumscribing the Brillouin zone Usually it is 4 6 The smaller bzr value the faster calculation the lower the accuracy These parameters should be the same as used by the NMTPLW calculation Finally path to additional input files must be specified gt fe hub ihub lt hubfile gt gt fe hop ihop lt hopfile gt gt fe opt iopt lt optfile gt fe enr ienr lt enrfile gt fe pnt ipnt lt pntfile gt e ihub lt hubfile gt this parameter has no effect in the linear response calcualtion e ihop lt hopfile gt this parameter has no effect in the linear response calcualtion 15 e iopt lt optfile gt this parameter has no effect in the linear response calcualtion e ipnt lt pntfile gt this is the file which contains a list of the wavevectors q for linear esponse calculations so called PNTFILE Use program QPNT located in maglib qpnt to prepare PNTFILE See section USING MAGLIB LIBRARY for the detailed description Lines in the PNTFILE numer
31. e same as the BNDFILE which contains the bands generated at the grid nk1 nk2 nk3 These files will be created automatically after one run of the main program For metals it is possible without lost of computer time to essentially improve accuracy of the BZ integration using multigrid technique The idea is based on the fact that the effects of energy bands and of the Fermi surface can be taken into account exactly in the linear response calculation While matrix elements necessary to construct induced charge density are calculated at the coarse grid set up by the numbers nk1 nk2 nk3 the weights for the k space integration which take into account the exact shape of the Fermi surface can be found using the bands generated at the dense grid set up by the numbers nw1 nw2 nw3 Two grids must commensurate with each other To reach this purpose first ENRFILE which contains the bands at the dense grid must be created To create ENRFILE just specify RUNMODE 0 The program will automatically generate k points according to nw1 nw2 nw9 will do only one band calculation will store ENRFILE and will stop Using any RUNMODE different from 0 will pick up coarse k space grid according to nk1 nk2 nk3 During the construction of the weights for the integration information stored in the ENRFILE will be taken into account In order to choose numbers of bands below and crossing the Fermi level use the following hint Look at the bands in the I point They usually
32. ly starting with the string wgt If key iwgt is set to zero WGTFILEs will be created as temporary and will be stored in the scratch directory e DRO contains changes in the induced charge density The file names will be constructed auto matically starting with the strings dro and dps Keys idro idps cannot be set to zero Other files are not used by the programs MAGASA MAGPLW Section OTHER INPUT DATA is also not used by the MAGASA MAGPLW 19 6 EXECUTING MAGASA PLW IN PARALLEL REGIME The linear response calculations for different q points give an opportunity to parallelize the execution i e submit jobs with different q points to different nodes It is also possible to parallelize the execution for different w and u It is not advised to split different G vectors since they are usually connected by symmetry Depending on how many nodes can be used for running the MAG different hints for parallelization can be given However in all cases one has to always keep in mind the following steps e i there is a number of files which must be prepared before running MAGASA PLW They include SCFFILE containing the charge density of the original crystal calculated using NM TASA PLW packages Main input control file INIFILE must be revised List of q points and G vectors must be prepared and stored in the PNTFILE The grid of wave vectors q must commensurate with the grids for k space integration and the grid for finding the integra
33. means that among a set of all q G points G is a reciprocal lattice vector the vector with minimal length will be picked out This option must be used when generating wave vectors q for linear response calculation 32 11 Acknowledgments I greatly acknowledge Dr Andrej Postnikov who has initiated writing of this manual Part of the developments has been done in collaboration with my brother Dr Dmitrij Savrasov Special thanks to Prof Ole Andersen and Dr Ove Jepsen who are my LMTO teachers 12 COPYRIGHT These programs are a free software for scientific and or educational purposes It is not allowed to redistribute them without prior written consent of the Copyright owners It is illegal to commercially distribute these programs as a whole or incorporate any part of it into a commercial product References 1 P Hohenberg and W Kohn Phys Rev 136 B864 1964 2 W Kohn and L J Sham Phys Rev 140 A1133 1965 3 For a review see also Theory of the Inhomogeneous Electron Gas edited by S Lundqvist and S H March Plenum New York 1983 4 E K U Gross and W Kohn Phys Rev Lett 1985 5 S Y Savrasov Phys Rev Lett 69 2819 1992 6 S Y Savrasov Phys Rev B 54 16470 1996 7 S Y Savrasov D Y Svarasov and O K Andersen Phys Rev Lett 72 372 1994 8 S Y Savrasov and D Y Savrasov Phys Rev B 54 16487 1996 9 S Y Savrasov Phys Rev Lett 81 2570 1998 33
34. ng for the scratch file names Scratch files will be created in the temporary directory The path to the temporary directory is contained in the file mag run setup f Always set iscr 0 ibnd lt bndfile gt this is the file which contains energy bands on the grid of k points in the IBZ Keep energy bands in the INP directory to make them shareable when running different q points spontaneously Always set ibnd 2 ipot lt potfile gt this file is not used by MAGASA PLW ifat lt fatfile gt this file is not used by MAGASA PLW idos lt dosfile gt this file is not used by MAGASA PLW iscf lt scffile gt this file is not used by MAGASA PLW Name for the SCFFILE which contains the self consistent charge density distribution for the original crystal as calculated by the program NMTASA PLW will be read from the terminal SCFFILE must be prepared before running linear response code For convenience place it into the INP directory tout lt outfile gt this is the current output file in this run In fact it is first character string of the filename The final name will also contain q point number as it was read from the Q POINT input line frequency name as read from FREQUEN input line and also task name as it was read from the RUNTASK input line Set iout 2 if storage is necessary set iout 6 to print the current output to the terminal Note that this file will be created NOT in the INP directory as it is done with CONFILE BNDFILE and PSIFI
35. ng spin susceptibility using the package MAGASA or MAGPLW For magnetic field calculation there is additional parameter w 1 0 1 fixing polarization of the field in spherical coordinates The main input file to the NMT pro gram INIFILE is used as the main input control file in the MAG Also a list of wave vectors q G must be prepared before running MAG and stored in the file called PNTFILE An other input control file called LRTFILE must be prepared typical example will be given The change in charge density for each q G w u will be calculated by MAG and stored in two kinds of files referred as DROFILE and DPSFILE DROFILE contains change in charge density expanded in spherical harmonics within MT spheres and DPSFILE contains change in charge density expanded in plane waves in the interstitial region There is as many DROFILEs and DPSFILEs as 3 x Ng x Natom All programs described in this manual can be downloaded from http www mpi stuttgart mpg de docs A About the notations in this document e all file names like nbc ini main exe are boldfaced e all directory names like magplw dat are italicized e capitalized names like INIFILE STRFILE are made to shorten references to the MAIN INPUT CONTROL FILE for INIFILE STRUCTURE CONTROL FILE for STRFILE etc I apologize if the description of some parameters is short and unclear Any suggestions to improve this manual are welcomed I also cannot guarantee that all the bugs in
36. onsider this example xxx Band Structure Calculation of bcc Fe x xx 3 lift 1 lmto 0 nsph ih lrwf 0 npfr Exchange Correlation Code 14 1 2 3 by VBarth H Gunn L Jan W Iterative Procedure Limits and Accuracy 50 0 2 0 2 1 D 4 6 0 0 3 0 3 niter mix mag eps lbr ibr mixb mixh Atomic Data 1 1 2 1 natom nsort nspin norbs 5 425 1 0 lattice parameter v v0 1 is iatom 3 0 05 0 0 1 0 0 0 2 5 0 0 Nkap Ekap ikap f t Ee a aa 26 D0 18 D0 1 1 0 0 5D0 55 847 z zcor lr icor ispl split mass 2 349 1 D0 1 D0 0 DO mt sphere rou sphere weight rloc 626 lmax t lmax b lmax v valence states are spdf states for E 0 5 Ry 4434 main quantum numbers 1110 basis set 3330 choice of Eny 0 30 0 30 0 30 0 50 Eny 0 30 1 60 1 50 3 00 Dny spdf states for E 0 5 Ry 4434 main quantum numbers 1110 basis set 3330 choice of Eny 0 30 0 30 0 30 0 50 Eny 0 30 1 60 1 50 3 00 Dny spdf states for E 0 5 Ry 4434 main quantum numbers 1110 basis set 3330 choice of Eny 0 30 0 30 0 30 0 50 Eny 0 30 1 60 1 50 3 00 Dny semicore states are 0 of states Input Control Files 2 gt 2 fe con icon lt confile gt O fe tmp itmp lt tmpfile gt 2 fe psi ipsi lt psifile gt O fe tmp itmp lt tmpfile gt 2 fe bnd ibnd lt bndfile gt O fe pot ipot lt potfile gt O fe ptn iptn lt ptnfile gt O
37. ormation below contains the Fermi energy found EF number of states calculated TOS which must coincide with the total valence charge DOS stands for the density of states Numbers of fully filled bands and the number of bands crossing the Fermi level are for information only They will not overwrite the input numbers nff nef from the INIFILE The line LR Information contains a hint for choosing the parameters nff nef in the linear response calculation Another hint is Locate the band nearest to the Fermi energy Step out from this band by approximately 0 5 Ry down Count the bands below this energy they can be treated as filled bands nff The bands located at the energy window plus minus 0 5 Ry from the Fermi energy should be treated as the bands crossing Ep nef Note that this is only necessary for metals use true number of filled bands and set nef 0 for semiconductors Also when calculating electron phonon matrices use true number of the bands crossing the Fermi level otherwise storage of the wave functions will be exceedingly large 25 kkkKKKKA BZINT started CPU time 63 59000 RR k k K KK K K K EF TOS DOS 1928632 7 989166 14 10590 EF TOS DOS 7936313 7 999986 14 09788 EF TOS DOS 7936323 8 000000 14 09784 EF TOS DOS 7936323 8 000000 14 09784 Calculated average square of electron velocities lt Vx 2 gt 7472877 lt Vy 2 gt 7472877 lt Vz 2 gt 6956963 Calculated bare plasma frequencies
38. point in order to make a perturbation vector q G Reciprocal lattice vectors are generated by shells in the reciprocal space and stored in the PNTFILE G SHELL 0 corresponds to G 0 then follows first second etc shells Setting G SHELL equal to 1 will consider a given q point plus all reciprocal lattice vectors G which belong to a given shell as perturbation vectors q G Some of these vectors can be connected by symmetry operations Note that GX 0 are only necessary to apply external fields changing within the unit cell All acoustic magnon modes for example are seen by considering only q with G 0 To find optical magnon modes it is necessary to apply non zero G FREQUEN gives the frequency in meV Note that FREQUEN must be set as a character string The reason is that all output files containing change in charge densities potentials etc must be named differently for different w The character string specifying w will be added to any output filename For example FREQUEN 0300 means that w 300 meV will be treated as the perturbation frequncy and the string 0300 will be added to any of the output files RUNMODE shows in which mode the program will be executed The following answers can be given 0 preparational mode Only the bands for a large number of k points will be calculated Use this mode when calculating metals For correct treatment of the effects of the Fermi surface the knowledge of the energy bands at a dense grid is
39. printed out in the OUTFILE Locate the band nearest to the Fermi energy Step out from this band by approximately 0 5 Ry down Count the bands below this energy they can be treated as filled bands nff The bands located at the energy window plus minus 0 5 Ry from the Fermi energy should be treated as the bands crossing Ep nef Set lorentizan broadening parameter pole to approximately 0 1 Ry 16 Note that any wave vector q must belong to the grid generated by nk1 nk2 nk3 and nw1 nw2 nw3 In other words grid of wave vectors q set up by three numbers nq1 nq2 nq3 see section USING MAGLIB LIBRARY must commensurate with the grids nk1 nk2 nk3 and nw1 nw2 nw3 Normally select such ng1 nq2 nqg when the number of irreducible q points generated is not more than 20 40 The grid for k space integration can be the same nk1 nk2 nk3 nq1 nq2 nq3 or not more than twice denser Do not use very many k points here all k space integrals are fastly convergent The grid nwi nw2 nw8 is the same as the grid nk1 nk2 nk3 for semiconductors For metals use nw1 nw2 nw3 which is 3 5 times denser than the grid nk1 nk2 nk3 The number of irreducible k points for integration weights should be of the order 1000 Final note must be said when calculating electron phonon matrices RUNMODE 5 Since the corresponding BZ integrals are very sensitive to the divisions nk1 nk2 nk3 use the following hints i Reach self consistency for all q points all displacements and all polariza
40. radius e rou sphere this parameter has no effect for linear response calculation using MAGASA PLW packages e weight this parameter has no effect for linear esponse calculation e rloc this parameter has no effect for linear response calculation 12 e Imaz t maximal angular momentum not l 1 for the basis functions i e for the decom position of the tails coming from other atoms Normally it is 6 for PLW calculation e Imaz b maximal l actually included in basis e Imazx v maximal l value for the expansion of the induced charge density and the induced poten tial in spherical harmonics Normally it is the same as Imaz t The following block is repeated for each from 1 to nkap valence states are n 4 5 Input Control Files COOWrFrFRFR HD oowre KH OOOO Rp pd 43 11 3 3 30 0 30 0 50 60 1 50 3 00 30 0 30 0 50 60 1 50 3 00 OOR MF OO OF MRF OOO Ah 0 30 0 30 0 50 1 60 1 50 3 00 states for E 0 5 Ry main quantum numbers basis set choice of Eny Eny Dny states for E 0 5 Ry main quantum numbers basis set choice of Eny Eny Dny states for E 0 5 Ry main quantum numbers basis set choice of Eny Eny Dny The meaning of these parameters is exactly the same as in the INIFILE of the NMT package Input Control Files 2 0 2 NNOOoOoNO fe fe fe fe fe fe fe fe fe con tmp psi tmp bnd pot ptn dos scf
41. re Calculation of E k with Eny Dny Cny Wny Et 500E 01 for Fe 45655 31805 84610 4 9488 for 4s state center 60984 51403 2 1312 1 9378 for 4p state center 65804 1 3297 76716 27252 for 3d state center Eny Dny Cny Wny Et 1 00 for Fe 45655 31805 84610 4 9488 for 4s state center 60984 51403 2 1312 1 9378 for 4p state center 65804 1 3297 76716 27252 for 3d state center Eny Dny Cny Wny Et 2 50 for Fe 45655 31805 84610 4 9488 for 4s state center 60984 51403 2 1312 1 9378 for 4p state center 65804 1 3297 76716 27252 for 3d state center 3kappa spin dn panel 1 Band Structure Calculation of E k with Eny Dny Cny Wny Et 500E 01 for Fe 45655 31805 84610 4 9488 for 4s state center 60984 51403 2 1312 1 9378 for 4p state center 65804 1 3297 76716 27252 for 3d state center Eny Dny Cny Wny Et 1 00 for Fe 45655 31805 84610 4 9488 for 4s state center 60984 51403 2 1312 1 9378 for 4p state center 65804 1 3297 76716 27252 for 3d state center Eny Dny Cny Wny Et 2 50 for Fe 45655 31805 84610 4 9488 for 4s state center 60984 51403 2 1312 1 9378 for 4p state center 65804 1 3297 76716 27252 for 3d state center doo BANDS finished CPU time 63 59000 aa k kk kkk 7 5 Integrating over Brillouin Zone After the band structure calculation finished the Brillouin zone integration starts controlled by BZINT see source file bzint f The inf
42. required This mode will prepare the necessary file 1 starting self consistency mode Use this mode when nothing is done with the self consistency for a given q point That means that the change in charge density is not known no DROFILE and DPSFILE exist 2 continuation of the self consistency Use this mode when self consistency for a given q point is continuing with the same set up of k points and other data were not changed DROFILE and DPSFILE must exist at this step but by some reason self consistency is not finished Actually if one sets RUNMODE 2 and the program will not find corresponding DROFILE and DPSFILE the mode will be automatically set to one RUNTASK selects polarizations m z p stand for applied magnetic fields polarized along 1 0 1 axis Note that the magnetization axis is always given by z axis Therefore in order to study transverse spin fluctuations in ferromagnets only m and p polarizations should be considered Another keyword v stands for applied scalar external potential in order to study charge response 4 MAIN CONTROL FILE INIFILE The main control file of the linear response package is very similar to that of the NUTASA PLW packages It has an extension ini Below an example of magnon spectrum calculation for Fe is considered The INIFILE for Fe has the name fe ini or more shortly ini It can be created easily from the INIFILE of the NMT package Only a few lines must be edited there Let s c
43. rge density The meaning of every parameter is given below e niter max number of iterations for doing self consistency of the change in charge density for given q G wp e mix starting mixing of the charge density in linear mixing scheme During the iterations towards self consistency the mixing will be optimally adjusted according to the Pratt scheme This parameter is ignored if Broyden mixing see below is switched on e mag starting mixing for the magnetization in linear mixing scheme During the iterations to wards self consistency the magnetization mixing will remain constant and will NOT be adjusted The parameter has no effect for non spin polarized calculations or if Broyden mixing see below is switched on e eps charge density convergency criterion Typically 107 The program will stop doing self consistency for a particular q G wp if the mean square difference between two induced densities in consequent iterations is less then eps Note that if a few displacements and polarizations run spontaneously the program will automatically terminate iterating those displacements which reach self consistency Also note that the reached convergency is written to the DROFILE When restarting the execution the program will check whether for a given displacement and polarization reached accuracy is higher than input eps or not If it is higher number of iterations for this displacement and polarization will be automatically set to 0 if it
44. ter ENRFILE is created Further execution depends whether one or more nodes will be used to run the program If only one node is used just select q point number G vector shell frequency w and polarization code and run the program with the RUNMODE 2 The CONFILE BNDFILE and PSIFILE will be automatically created and placed into the INP directory WGTFILE for this q point and w will be also automatically created and placed into WGT directory After the execution is completed DROFILE and DPSFILE for this q G w y will be created and placed into the DRO directory If it is possible to use several nodes one should first submit a job for a particular q G w u and wait until CONFILE BNDFILE and PSIFILE will be created After that it is allowed to submit another q G w y point for a different node since this execution will use CONFILE BNDFILE and PSIFILE created before 20 7 OUTPUT MESSAGE FILE OUTFILE Here the description of the output messages is given Also short introduction to the structure of the program is described see Figure 2 Consider the output file made for NbC during the calculation of the dynamical matrix RUN MODE 4 for the point q 0 0 1 Running the program in other modes produces similar output 7 1 Reading Input Data The execution of the MAGPLW package source file magmain f starts from reading the input data controlled by INIT subroutine see file init f Beginning of the OUTFILE contains the information read
45. the programs are found Since linear response calculation is a relatively new field in computational solid state physics there could be some other problems instabilities in the algorithms which have been used in the programs Any further improvements of the programs are welcomed 2 INSTALLATION In this section the directories which are used for running the programs and storing input output data are described The MAG packages have the directory organization similar to the packages NMT There exists three main directories e magasa magplw directories containing source code of the main LR LMTO programs MA GASA MAGPLW designed to perform self consistent linear response calculations Input data files and some auxiliary programs are also stored in this directory e maglib directory containing a number of other application programs helping to construct input data and understand output information Below the contents of each of the directories is described The magasa and magplw have two subdirectories e magasa run magplw run directories containing the source code f of the main LR LMTO programs MAGASA MAGPLW text of the program written on FORTRAN77 object files o and executable file usually named as main exe e magasa dat magplw dat directories with the input output data files to the main LR LMTO program Usually many subdirectories are created here according to the element com pound name to be calculate
46. tion weights Place all these files into the INP directory Place also here the STRFILE which is the control structure file of the NMT package e ii there is a number of files which is prepared by the program MAGASA PLW at the beginning these files are unique for all q points and can be used as shareable These files include ENRFILE with the energy bands at the dense grid nw1 nw2 nw3 which will be used to find integration weights CONFILE BNDFILE and PSIFILE are the files which contain structure constants energy bands and wave functions at the coarse grid given by nk1 nk2 nk3 When running the MAGASA PLW checks whether the file exists or not and if the file does not exist it will be created automatically e iii there is a file which depend on the q point and frequency but does not depend on G and u This is WGTFILE which contains k space integration weights WGTFILEs are kept in the WGT directory When running the MAGASA PLW checks whether the file exists or not and if the file does not exist it will be created automatically According to these steps it is first of all necessary to prepare ENRFILE with the energy bands for the integration weights Skip this step if the system is a semiconductor In order to prepare ENRFILE run the main program with the RUNMODE 0 The number of k points generated in this case will be equialent to the dense grid controlled by parameters nw1 nw2 nw3 of the INIFILE The program will automatically stop af
47. tions All DROFILEs and DPSFILEs must be created and stored in the DRO directory Calculate dynamical matrix and the phonon spectrum for every q point when for this q point self consistency is reached for all displacements and all polarizations ii Delete contents of the directory WGT which contains the weights for the BZ integration iii Make a new INIFILE in which specify the divisions nk1 nk2 nk3 of the coarse grid equal to the divisions nw1 nw2 nw8 of the dense grid the number of irreducible k points for integration must be of the order 1000 Carefully set the number of bands below crossing the Fermi level equal to the actual number of bands below crossing the Fermi level Delete or rename BNDFILE and CONFILE since they contains the information generated for the coarse grid iv Run the main program with the RUNMODE 5 CONFILE and BNDFILE will be created Also a new PSIFILE with the wave functions will be created However the wave functions will be stored only for the bands crossing the Fermi level that s why it is necessary to reset this number otherwise this file will have a huge size v After CONFILE BNDFILE and PSIFILE are created at the dense grid nk1 nk2 nk3 nw1 nw2 nw8 run different q points to calculate the electron phonon matrices The new weight files will be created automatically and stored in the WGT directory Note that the integration weights for the electron phonon matrices are different from those which have been using
48. to find the induced charge densities that s why it is necessary to delete the contents of the directory WGT before using RUNMODE 5 The calculation of the electron phonon matrix is computationally much more faster than making self consistency and finding dynamical matrix 17 5 LINEAR RESPONSE CONTROL FILE LRTFILE LRTFILE is used to set up the data which are specific for linear response calculations In fact this file does not contain an information on the compound to be calculated and therefore the LRTFILE does not have to be modified except some special cases A typical LRTFILE is given below SET UP DATA FOR LINEAR RESPONSE CALCULATIONS GENERAL SETTINGS gt Magnons 7 Response scheme Phonons Magnons gt Plz off Polarizabilty on off Stn off 4 Stoner renormalization on off Del on Changes in radial functions on off Dyn none Dynamical matrix scheme none hf okal okax DEFAULT FILE SETTINGS O DSF dsf idsf lt dsffile gt 2 WGT wgt iwgt lt wgtfile gt O TMP tmp itmp lt tmpfile gt O STN stn istn lt stnfile gt O PHN phn iphn lt stnfile gt O PLZ plz iplz lt plzfile gt O PLZ pls ipls lt plsfile gt O POT dsv idsv lt dsvfile gt O POT dpv idpv lt dpvfile gt 2 DRO dro idro lt drofile gt 2 DRO dps idps lt dpsfile gt OTHER INPUT DATA 0 0DO 0 03D0 30 Wmin Wmax Nomg There several sections in it
49. ttice vector G and the frequency w the problem is reduced to the self consistent solution of a differential equation so called Sternheimer s equation which is a linearized version of Schr dinger s equation Solving this problem self consistently gives an access to the function x r q G w which is the susceptibility Fourier transformed over one index The entire problem is very similar to the self consistency problem of standard band structure calculation It is however much more heavy to solve it numerically because finding x r q G w for a grid of wave vectors q G requires Nyya total number of q G points self consistent calculations performed for a set of frequencies w It therefore requires N G X No self consistent calculations each of them being equivalent to the standard band structure calculation for the unperturbed system Linear response calculations of x r q G w should in principle not be so sensitive to the details of the charge density distribution over the cell However a full potential FP solution of the problem is desired According to different treatment of the full potential terms in the band structure calculations performed with the original FP LMTO method for the description of these programs see FULL POTENTIAL LMTO PROGRAMS NMT USER s MANUAL which is available on the WEB site under the address http www mpi stuttgart mpg de docs ANDERSEN there are two linear response programs MAGASA and MAGPLW
Download Pdf Manuals
Related Search