The User Manual
Contents
1. sees enne nnne 49 4 7 Compatible runs with PWSCF essere netten enne 50 5 Pre and post processing programs eeseeeeeeseeeeeeenee nennen eene ren eene nennen 52 5 1 Pre processilg ree era ee er me ee m e e e denies 52 5 1 1 convert to Confis Xeen ela aei aae n eee one de dee Re enn d 52 5 1 2 de c aesareeediseaisi 52 5 2 Post processing eene merce e teer ete tope ee tne e De RE gts 53 5 2 1 convert from config x sssssssssesseseseeeee enne eene nnne entr ennns einen 53 5 2 2 plot band structure x esses enne nennen enne 54 5 2 3 plot d0s X see RR e e e eR Re nice 54 5 2 4 Convert TOK is esos rescate ie De Gate Caere aS Dr CE CEN VER TM 55 Appendiks Dm 56 Appendix B uu Mete etie e tes 57 1 Introduction PWmat package is a plane wave pseudopotential package for density functional theory DFT calculations It can perform the following calculations indicated by the JOB flag in etotinput SCF self consistent field calculation NONSCF non self consistent field calculations e g for band structure calculation DOS density of state calculation which is usually after a NONSCF or SCF calculation RELAX atomic relaxation calculation MD ab initio molecular dynamics calculation and NEB nudged elastic band calculation for barrier heights PWmat can be run on CPU or CPU GPU processors In the GPU2. MP N123 nkl nk2 nk3 skl sk2 sk3 run gt check x to generate IN KPT then run a SCF calculation get OUT VR and copy OUT VR to IN VR After that set KP INV in etot input comment out the MP NI23 line using gt check x to generate a different IN KPT note it will overwrite the previous IN KPT which contains all the k points in a band structure path for the typical primary cell shapes one can also edit IN KPT by hand then run a NONSCF calculation using this IN KPT IN KPT T IN VR T The band structure information will then be reported in REPORT One can use plot band structure x for post process to view the band structure 45 4 3 Density of States Calculations JOB DOS This usually also follows from one SCF calculation just like for NONSCFP then to calculate the DOS in this step In other words there need to have three calculations in order to get DOS As for NONSCE one first gets OUT VR from SCF calculation Then in DOS calculation copy OUT VR as IN VR set IN VR T read this IN VR Also one might want to use more k points for a nice DOS To do that one can use a larger Monkhorst Pack grids in etot input MP N123 nk1 nk2 nk3 sk1 sk2 sk3 use gt check x to generate a new IN KPT Then one needs to do a JOB NONSCF calculation to get the eigen energies stored in EIGEN ALL and eigen wave functions OUT WG If one doesn t want to use more k points then one can use the SCF
3. cudazone nvidia cn 57 58
4. The PWmat will automatically detect whether there is a velocity section in atom config If yes then it will use it as the initial velocity If no it will use templ to randomly 49 generate an initial velocity When using JOB MD one has to include an etot input line MD DETAIL md mstep dt templ temp2 md specifies the method For md 1 verlet only when it is start from scratch templ will be used It is used to generate the initial random velocity according to this temperature For md 1 temp2 is not used For continued run the initial velocity is read in from the atom config file so templ is not used For md 2 3 if start from scratch initial atom config does not provide the velocity then templ is used to generate the initial velocity The md 2 3 dynamics will scale the temperature linearly with steps from temp to temp2 4 7 Compatible runs with PWSCF The PWmat can be run compatibly with the open source code PWSCF Mostly the PWmat can generate the wave function charge density and potential files which can be read by the PWSCF program or the PWSCF compatible programs e g Wannier90 function generator or GW calculations These programs can be run on CPU To run those programs the user is responsible to prepare their control input files One should also copy the corresponding PWSCF pseudopotential files from our library they are the same vwr atom uspp atom txt potentials but in the PWSCF pseu
5. The default values are determined by Ecut2 i e make sure the Ecut2 sphere can be held inside the nl n2 n3 reciprocal box NS123 The format is like this NS123 nls n2s n3s nls n2s and n3s are the real space FFT grid point to calculate the real space projector function So these are only used for NONLOCAL 2 For small systems nls n2s n3s are the same as nl n2 n3 For large systems smaller values can be used to save time for projector generation N123L The format is like this N123L n1L n2L n3L nlL n2L and n3L are the real space grid for hard charge density The default values are determined by Ecut2L For norm conserving pseudopotential the soft charge equals hard charge Ecut2L Ecut2 so n1L n2L n3L equal n1 n2 n3 For ultra soft Ecut2L 4 Ecut2 n1L n2L n3L 2 nl 2 n2 2 n3 SPIN Control the spin of the DFT calculation SPIN 1 no spin the default SPIN 2 with spin up and down but not non collinear spin There is no spin orbit coupling in this version of PWmat 14 XcFunctional Control whether to use GGA calculation xcfunctional 1 using PBE GGA xcfunctional 0 use LDA the default COULOMB Control the Poisson equation solution for the Coulomb interaction COULOMB 0 the periodic boundary condition the default COULOMB 1 x1 x2 x3 the isolated cluster boundary condition It can avoid the image interaction in this calculation Th
6. e g 4 lines see below while other PW calculation parameters will be provided by default However by running check x it also generates a long version etot input long which can be copied and used as etot input with manual changed parameters for advanced users 27 2 IN ATOM atom config JOB SCF IN PSPI Si ncpp upf We suggest the following specific steps to run PWmat from the beginning to the end 1 Prepare atomic position file atom config from online database or and visualization packages e g VESTA Avogadro The output system file format from these packages could be system xyz system xsf or system vasp 2 Convert the system xyz system xsf system vasp format into atom config by running our serial utility code gt convert_to_config x lt system xyz xsf vasp It will generate atom config file Note atom config can also be prepared by hand or 2 user s own software 3 Prepare etot input see detail explanation in the next section 4 Pre process by running gt check x This will tell you whether there is any error or inconsistency in etot input It will also generate the real space grid nl n2 n3 This will help you to decide the nodel node2 the first line in etot input One needs to make sure that n1 n2 can be divided by nodel Thus one might need to change nodel in etot input according to nl n2 n3 or increase nl n2 n3 Note the total number of CPU GPU equals nodes1 no
7. kptl kpt2 spins ispinl ispin2 bands iwl iw2 in the file OUT RHO SP This is different from OUT RHO since it can select which wave function to be included in the charge density For Iflag 0 default without output Iflag 1 output partial charge at the end of other 17 calculations Iflag 11 output partial charge using initial input wave functions before doing any other calculations then stop the code Iflag 2 output the wave function instead of the charge density at the end of other calculations Iflag 22 output the wave function using initial wave functions before doing any other calculations then stop the code For iflag 1 11 dens r i A r where index i is defined by the above mentioned intervals For iflag 2 22 the wave functions are stored one after another in OUT RHO_ SP in the sequence do ik kptl kpt2 do is ispinl ispin2 do iw iwl iw2 real part do ir 1 n Real psi ir Imag part do ir 1 n Imag psi ir 5 5 OUT FORCE OUT FORCE T the PWmat will calculate the atomic force and output the force in file OUT FORCE This is for one shot one atomic position snap shot JOB SCF calculation only For JOB RELAX or JOB MD PWmat will always calculate the force Also note this will not work for JOBZNONSCT since there the total energy and SCF charge density will not be calculated OUT FORCE
8. the atomic configuration of the second minimum for NEB algorithm while the atom config in the IN ATOM atom config contains the first minimum atomic configuration and Nimage image configurations points will be created between these two fixed atomic configurations Note the atom config and atom2 config must have the same atom orders If itype_at2 2 the atom config in 29 IN ATOM atom config is no longer used although it still need to be presented in etot input and the atom2 config must contains Nimage 2 configurations two ends plus Nimage images Besides it must also has a fixed form for each configurations Natom Lattice Position Force line In practice this atom2 config is often copied from MOVEMENT from previous NEB runs 2 3 Pseudo potential files atom UPF They are the files copied from the provided libraries The default Ecut wfc_cuttoff is provided in those files They are ASCII files thus can be read In the execution of PWmat the corresponding pseudopotential files indicated in etot input must be copied to the running directory The UPF is the format developed in Quantum Espresso Currently PWmat can use norm conserving pseudopotential or ultrasoft pseudopotential There are many groups developing pseudopotentials and most of them are in the UPF format 2 4 IN KPT The file IN KPT contains the k point vectors and their weights It can be generated from preprocessing by running gt check
9. wavefunctions into num blocked psi parts and then put the parts into GPU memory successively one after another during scf iteration This is to save the use of GPU memory If a previous run found the GPU out of memory this can be tried Larger the num blocked psi smaller is the GPU memory used Num blocked psi l default the program will try to detect the memory situation automatically If it estimates that there is not enough GPU memory PW mat will automatically split the wavefunctions into multiple blocks and to send each block to GPU memory successively This parameter intends to save the GPU memory to calculate a larger or more complicated systems So when PWmat tells 26 CUDA Memory insufficent one can try this parameter by setting num blocked psi 2 3 or 4 and etc Note that using this parameter will reduce the speed of PWmat e g by a factor of 1 5 Write2Memory Write2memory 0 write the wavefunctions into disk Write2memory default write the wavefunctions into cpu memory This parameter is used to save cpu memory to calculate a larger or more complicated systems in particular for the case multiple k points are calculated then one can use Write2memory 0 Note that it will reduce the performance of PWmat in some degree 2 2 atom config This file describes the supercell box atomic positions optionally the atomic force and atomic velocity of the system It h
10. will not calculate total energy but it can be used to calculate band structure or density of state JOB RELAX do atomic position relaxations using DFT force and total energy Each atomic relaxation step will do one SCF calculation Might want to have RELAX DETAIL in the etot input See below A concise result will be reported in RELAXSTEPS The atomic movements for each relaxation step are reported in MOVEMENT JOB NEB calculate the barrier height using nudged elastic band method Must have a variable NEB DETAIL in etot input file Besides IN ATOM which gives the first valley site atomic position there must be a second valley site position given in the NEB_DETAIL line One must precalculate e g using JOB RELAX the atomic configuration of these two valley sites before using JOB NEB to calculate their barrier See NEB_DETAIL for more details Output files RELAXSTEPS NEB BARRIER MOVEMENT NEB BARRIER gives the barrier height information JOB DOS do density of state DOS calculations based on wave function and eigen energy input from previous calculations must have IN WG zT which is also default and EIGEN ALL from previous step calculations Output files DOS totalspin bpsiiofil1 0000x It takes the converged wave function IN WG do an atomic orbital projection in order to get the partial DOS JOB MD do molecular dynamics MD simulations Must have variable MD DETAIL in etot inpu
11. 0 10000 1 SCF ITER1 6 4 ALGORITHM1 3 1 0000 0 10000 1 3 1 0000 0 10000 1 3 1 0000 0 10000 1 3 1 0000 0 10000 1 3 1 0000 0 10000 1 3 1 0000 0 10000 1 NONLOCAL 2 Rcut 3 20000004768372 Kok K K K K K K oe ok K K ok ok K oe ok oe oe ok K K K K oe K K K oe oko K K K K K K K K K K K K K K M xoeeeeeE end of etot input report the above are etot input long can be copied into etot input kk ok ok ok ok ok K oe ok K oe K ok ook ok oe oe ok K oe K ok K K ok oe oe K K oe oe ok oe oe K K e K K e K K K OUTPUT FILE FROM PETOT for more inf see file etot input atom config file 0 000000000000000E 000 FKK K ok ok K K K K K K K K K K K K K K K ke K K 36 recommended n1 n2 n3 from Ecut 59 8 59 8 59 8 recommended n1 n2 n3 from Ecut2 42 3 42 3 423 Actual ni n2 n3 used here 45 45 45 FFT grid determined by Ecut2 recomm niL n2L n3Lfrom Ecut2L 846 846 84 6 Actual n1L n2L n3L used here 90 90 90 double grid for charge density for uspp KKK ook K K K oe K oe oe K K K K K K K K K K K K nnodes_b 5 num_group_b 1 num_group_k 1 nnodes_b node1 num_group_k node2 num_group_b 1 natom 64 ipsp all 0 ipsp_all 0 vwr psp 1 ultrasoft psp islda 1 igga 0 0000000000 Ecut 21 00 Ecut2 42 00 Ecut2L 168 00 Smth 1 00 Smth always 1 in this code Ecut in units of Ryd totNel 570 00 mx 312 tolug 1 0E 03 tolE 1 0E 05 totNel num of electron mx the num of calc
12. 0 517507E 00 0 883061E 00 5 0 75234053674623E 03 0 512514E 00 0 961894E 00 6 0 75234044035416E 03 0 517438E 00 0 961928E 00 7 0 75266732167841E 03 0 526413E 00 0 883176E 00 8 0 75296021410969E 03 0 530651E 00 0 356206E 00 9 0 75304050727950E 03 0 520291E 00 0 846314E 00 10 0 75305820226225E 03 0 504589E 00 0 944466E 00 11 0 75306185743092E 03 0 000000E 00 0 000000E 00 This means Nimage 10 The Etot eV indicates the total energy of this image Dist is the distance between the neighboring images between image and image 1 For good NEB run the distance should be roughly the same Angle is the cos theta of the angle theta between two R image 1 R image and R image R image 1 For good NEB run cos theta should be close to 1 especially around the barrier height In practice it should be fine as long as the cos theta is close to 1 near the barrier height Also for a good NEB run the distance between the images should be roughly equal Iter 19 means this is the 19 line minization result In NEB BARRIER it writes out the results for every relaxation iterations 3 5 MDSTEPS The MDSTEPS is the file concisely describes the steps of a molecular dynamics simulation It can have the following format Iteration 8 Etot Ep Ek 0 2346080032E 04 0 2346382949E 04 0 3029172416E 00 Temperature 996 40333 dE 22E 05 dRho 0 18E 03 SCF 2 Iteration 9 Etot Ep Ek 0 2346080014E 04 0 2346401749E 04 0 3217343276E 00
13. JOB DOS for which the default value is 3 8 Larger the Rcut more accurate but more expensive QijL0 GS QijLO0 GS 1 or 2 This parameter is only used by ultra soft pseudopotential calculations It controls the core charge implementation in ultra soft pseudopotential 1 means the s component of the core Qij charge is used and implemented in G space 2 means the s component of the core Qij charge is implemented in real space The default is 1 We strongly recommend the use of default value since QijLO_GS 2 might introduce some jitter in the energy curve QIJ PD QIJ PD 0 or 1 It controls whether to include the l p and d angular momentum core charge density QIJ r l in the ultrasoft 25 pseudopotential calculation The default is 0 which means only the s angular momentum core charge density is used We strongly recommend the use of 0 The difference is usually very small but the including of p and d components of the QIJ r can make the energy manifold unsmooth thus makes the JOB RELAX difficult PWSCF PWSCF T or F This parameter controls whether to output pwscf compatible output files wave function charge density and potential so the result can be run subsequently on pwscf For T it will output those files For F it will not output The default is F Num blocked psi num blocked psi 1 2 3 4 5 In choosing this parameter not 1 PWmat will divide the
14. K K oe oe K oe oe KK K K K K K K within atom mov step 1 Oe oe ok ok K K ok K oe ok ok oe K ok K oe ok oe oe ok oe oe ok ok K K ok K oe K K K K K K oe K K K iter 1 avelin 4 0 iCGmth 3 Ef eV 0 5228815E 01 err of ug 0 1190E 00 dv_ave drho_tot 0 1479E 00 0 4236E 01 E_tot 23417524427761E 04 2342E 04 mch_pulay drho_in drho_out 0 3742E 00 0 2839E 00 3 3RELAXSTEPS 38 The relaxsteps is the file concisely reports the atomic relaxation steps and the energy and atomic forces at each atom as well as the SCF convergence for each ab initio step calculations A typical RELAXSTEPS file will look like It 0 NEW E 0 7526919500493E 03 Av_F 0 17E 00 M_F 0 32E 00 dE 4E 04 dRho 4E 03 SCF 4 dL 70E 01 p F 0 38E 01 p F0z 0 77E 01 Fch 0 10E 01 I 1 CORR E 0 7527130487491E 03 Av_F 0 18E 00 M_F 0 37E 00 dE 3E 04 dRho 2E 03 SCF 3 dL 14E 00 p F 0 23E 02 p F0z 0 77E 01 Fch 0 10E 01 It 2 NEW E 0 7527421363137E 03 Av_F 0 10E 00 M_F 0 20E 00 dE 5E 04 dRho 9E 03 SCF 2 dL 0 49E 01 p F 0 19E 01 p FO 0 51E 01 Fch 0 10E 01 It 3 CORR E 0 7527473988358E 03 Av_F 0 12E 00 M_F 0 23E 00 dE 7E 05 dRho 3E 03 SCF 2 dL 0 78E 01 p F 0 80E 03 p F0z 0 51E 01 Fchz 0 10E 01 It The index of total line minimization number iteration or step index NEW this is a new line minimization direction The search direction p has changed CORR this is a middle step in the line minimization process corre
15. Pay attention to w_scf This parameter is used to stop the SCF iterations It takes the last iteration average force and the estimated force error estimated from V in V outl in SCF calculations if the estimated force error is less than last iteration force w scf then it will jump out the SCF loop So smaller w_scf and larger niter1 more SCF loop might be carried out and more accurate will be the force This might be particularly critical if very accurate final atomic positions are needed The default w scf for RELAX is 0 003 There are two relaxation methods imth 1 conjugated gradient method imth 2 BFGS method the imth is specified in line RELAX DETAIL Their performances are similar More advanced methods will be introduced in later version of PWmat When JOB RELAX is used one can also include an etot input line RELAX DETAIL imth nstep force tol If the max force becomes smaller than force tol a u the relaxation step will stop before nstep 47 4 5 Nudged Elastic Band Calculations JOB NEB Nudged Elastic Band NEB method is often used to calculate the potential barrier from one local minimum configuration to another local minimum configuration In order to do NEB calculation the two local minimum must be known already They can be calculated by JOB RELAX with their atomic configurations being atoml config atom2 config and energies being EI E2 The idea of NEB is to use a string of images configuration poi
16. Temperature 1058 29947 dE 55E 05 dRho 0 16E 03 SCF 2 Etot is the total energy DFT energy plus kinetic energy in eV Ep is the potential energy here DFT energy in eV Ek is the kinetic energy in eV For a Verlet algorithm when everything run well Etot should be an constant 41 Temperature is calculated from the Ek in Kelvin dE is the E n E n 1 in the SCF iteration eV drho lrho in rho outl relative error in the SCF calculation SCF is the number of SCF iterations for this MD step 3 6 MOVEMENT This is the file generated in RELAX NEB and MD It outputs the atom config of every atomic movement steps including the correction steps in the line minimization of RELAX in a single file one after another It contains the atomic position atomic force sections For MD it also contains the velocity section So it can be copied to atom config to continue the run of MD It can also be converted to other format for visualization e g as animation by using convert from config x lt MOVEMENT For JOB NEB the MOVEMENT contains the configurations for all the image points Note inside MOVEMENT the new configuration is appended on the old ones already in the file The format of MOVEMENT is the same as in atom config 3 7 Other output files There could be other output files e g OUT WG wave function output file OUT RHO charge density output file OUT VR potential output file OUT DENS selective w
17. calculation s eigen energies and wave functions thus skip the JOB NONSCF calculation step After this one can copy OUT WG to IN WG and do a JOB DOS calculation The PWmat will output a file DOS totalspin if SPIN 2 there will also be DOS spinup DOS spindown The format in DOS totalspin is each line example Energy Total Zn s Zn p Zn d O s O p O d One can plot this file for graphics The default energy smoothing broadening parameter is 0 1 eV If one wants to have different broadening parameter or have partial DOS for different atoms one can use the postprocessing utility code plot_dos x In order to have atom selective partial DOS one needs to provide a modified atom config file with the position section looks like 30 0 952534560 0 363594470 0 382027650 1 1 1 wl 30 0 540553000 0 850230410 0 966359450 1 11 w2 46 16 0 242857140 0 140553000 0 68433 1800 1 1 1 w3 Here wl w2 w3 are the weights for this atom in the partial DOS The plot dos x also uses bpsiiofill0000x which are the eigen wavefunciton to atomic orbital projection coefficients for different k points x which is output from the JOB DOS run 4 4Atomic Relaxation JOB RELAX This is to relax the atomic positions following the DFT energy and force It will generate the RELAXSTEPS and MOVEMENT files If it is not fully relaxed the MOVEMENT can be copied to atom config remove all the other iterations except the last one then run the PWmat again
18. k points are defined as k G1 ak1 G2 ak2 G3 ak3 Here G1 G2 G3 are the reciprocal lattice vector of lattice AL 3 3 Weight This is the weight of this reduced k point it can represent several symmetric k points This is used for SCF calculations and the total weight as the sum of individual k points should be 1 2 5 IN SYMM This is the symmetry operation file usually generated by check x It contains the space group It has the following format 12 24 nsym nrot identity and corresponding fractional translation 1 0 0 0 1 0 0 0 1 0 000 0 000 0 000 180 deg rotation cart axis 0 0 1 and corresponding fractional translation 1 0 0 0 1 0 0 0 1 0 000 0 000 0 500 180 deg rotation cryst axis 1 1 0 1 1 1 0 1 0 0 0 1 32 39 66 The first line is the two variables nsym and nrot nsym is the number of the crystal symmetries operations space group and nrot is the number of the crystal Bravais lattice symmetries only for the lattice not considering the atoms thus nrot is always larger than nsym For PWmat only nsym is used For the rest of the file there will be nsym operations each has this following format 180 deg rotation cart axis 0 0 1 and corresponding fractional translation E 0 0 s L 1 8 2 Is 3 I 0 0 s 1 2 s 2 2 5 3 2 0 0 1 s 1 3 8 2 3 5 3 3 0 000 0 000 0 500 11 1 1 2 1 3 The f
19. number of this atom x1 is the fractional coordinate of this atom in the unit of AL 1 x2 x3 are the fractional coordinates of this atom in the units of AL 2 ALC 3 Imvl imv2 imv3 indicates in the atomic relaxation whether this atom will move in x y z note not the x1 x2 x3 directions Imv1 2 3 1 move 0 not move Note the x y z coordinates of this atom can be calculated as X AL 1 1 x1 c AL 1 2 x2 AL 1 3 x3 28 Y ALQ 1 x1 AL 2 2 x2 AL 2 3 x3 Z AL 3 1 x1 AL 3 2 x2 AL 3 3 x3 Force The header of the force section This section is optional It will be followed by natom lines in the following form zatom fx fy fz 30 0 0372 0 01112 0 1021 They are the x y z direction atomic forces in eV Angstrom Velocity The header of the velocity section This section is optional If will be followed by natom lines in the following form zatom V X V y VZ 30 0 39292 0 222933 0 28211 They are the x y z direction atomic velocity in Bohr fs Note the atom config file can be converted from other format xyz xsf file using convert to config lt system xyz or xsf It will generate a system config file The atom config file can also be converted into xyz xsf file using gt convert from config lt xatom config It will generate the atom xyz and atom xsf files Under JOB NEB in NEB DETAIL line if itype_at2 1 then atom2 config contains
20. run one CPU is bundled with one GPU The package also come with a well tested pseudopotential library in the upf unified pseudopotential format for norm conserving pseudopotentials atom ncpp upf and ultrasoft pseudopotentials atom uspp upf PWmat is fully compatible with the popular open source code Quantum Espresso module pwscf with exchangeable data files but it can be an order of magnitude faster than pwscf since it runs on GPU In order to run PWmat one needs to provide the following input files in the running directory etot input the PWmat control file atom config the atomic position file atom upf pseudopotential files they can be copied from the provided pseudopotential library Optionally one might also need to provide IN KPT k point file IN SSYMM symmetry file IN VR potential input file IN WG wave function input file and IN RHO charge density input file These files can either be generated from the pre processing code check x e g for IN KPT and IN SYMM or from previous calculations IN VR IN WG IN RHO There are three steps to use PWmat 1 Prepare atom config and etot input and copy atom upf 2 Run PWmat 3 Post process visualization We provide suggestions for how to generate atom config e g through open source visualization tools One can also write atom config by hand The etot input which tells PWmat what to do can be very simple consisted with only a few lines
21. the line minimization in the total steps Force tol in eV A is the force tolerance for the maximal residual force If the maximum force is less than force tol the relaxation will stop The default values are RELAX DETAIL 1 100 0 01 Note the JOB RELAX will output a RELAXSTEPS and MOVEMENT output files While RELAXSTEPS gives a summary of the steps MOVEMENT records the atomic positions for all the steps Ecut The plane wave cutoff energy for wavefunction in Ryd note 1 Ryd 13 6057 eV The default value of Ecut is taken from the pseudopotential files atom upf from its wfc cutoff value Ecut2 The cutoff energy for the soft charge density and the potential in Ryd Ideally for high accurate calculations Ecut2 should equal 4 Ecut But in reality smaller Ecut2 can be used e g 3 Ecut or 2 Ecut By default Ecut2 2 Ecut The nl n2 n3 are determined by Ecut2 Note the rho cutoff value in the pseudopotential files atom upf is not used Ecut2L The cutoff energy for the hard charge density in Ryd Usually Ecut2L Ecut2 for norm conserving pseudopotentials and Ecut2L 4 Ecut2 for ultra soft pseudopotentials Note the rho cutoff value in the pseudopotential files atom upf is not used N123 The format is like this N123 n1 n2 n3 13 nl n2 and n3 are the real space grid to describe the wave function or soft charge density in real space It is also the FFT grid
22. x In gt check x it will use the Monkhorst Pack line MP N123 nk1 nk2 nk3 skl sk2 sk3 in etot input and symmetry of the system to produce the irreducible k points for total energy calculations The IN KPT for band structure calculation can also be produced by running gt check x The check x will automatically read KP INV in etot input and then generates the IN KPT 30 Note that for JOBZNEB calculation very often the image configuration along the path might have lower symmetry than the one at the initial minimum in IN ATOM atom config Since symmetry is generated based on atom config one must use a general not highly symmetric atom config when generating IN SYMM and IN KPT for NEB calculations Finally the IN KPT can also be edited by hand It has the following format 2 nkpt 2 1 0000 iflag a0 0 250 0 250 0 250 0 25 aki ak2 ak3 weight 0 250 0 250 0 750 0 75 Variable Specification nkpt The number of k points iflag iflag 1 the k points are in x y z directions which is defined by the x y zin AL 3 3 in atom input iflag 2 the k points are in the reciprocal lattice of the super cell AL 3 3 a0 will not be used a0 only used when iflag 1 in atomic unit Bohr akl ak2 ak3 iflag 1 the k points are defined as kpx 2 zx akl a0 kpy 22 x ak2 a0 31 kpz 2 x ak3 a0 iflag 2 the
23. 20 4 ALGORITHMO zn 0 0000 0 10000 1 3 0 0000 0 10000 1 3 0 0000 0 10000 1 3 1 0000 0 10000 1 Imth icmix dE Fermi Dirac 4 2 None self consist calculations JOB NONSCF This is usually used following a SCF calculation to study the electronic structure of the system in particular the band structure It can use the OUT VR from the previous calculation copy them to IN VR and set IN VR to T in etot input but with different k points in IN KPT to study the band structure It can also be used to study some 44 special cases e g with patched together potential or charge density Note in NONSCE it can still generate the potential from an input charge density One needs to set IN VR T to input potential from IN VR or IN RHO T to input charge density from IN RHO then generate the potential It will then simply calculate the eigen energies e g for different k points listed in IN KPT The true differences between JOB SCF and JOB NONSCF is at the line SCF ITERO 20 4 ALGORITHMO 3 0 0000 0 10000 1 3 0 0000 0 10000 1 3 0 0000 0 10000 1 3 0 0000 0 10000 1 Imth icmix dE Fermi Dirac In SCF the icmix for the late SCF iteration steps must be 1 Indicating there will be charge mixing and charge update while for NONSCE all the icmix will be zero no charge mixing and charge update To do a band structure calculation it usually run PWmat in the following procedure Set Monkhorst Pack line in etot input
24. 2VMAI The User Manual August 14 2015 Beijing LongXun Quantum Technology Co Ltd I Introduction eere ee RETE HRO e RSEN E RO REN E E IRE reset E 1 2 put eS p 5 2 1 ClOLMPUl MEME c iia abel Mae ee dale E A ely 5 2 2 atomiconfig gone gue dee ale ashes ORE RP ERR Dies 27 2 3 Pseudo potential files atom UPF esee 30 Pr EM LN Cd ud Up AE 30 25 EIL M 32 2 6 Other mput files cce cn asst Aes e Re RC RECHERCHER es 33 3 Output files oett Et et te to Eti Pe e Et nie e Er Ene eect 35 31 Standard output file certet em ee 35 EB REPORT EE 35 33 RELAXSIEPS edskessssd Hes dU ROUTE REED te ER SS 38 34 NEB BARRIER o minero RHENO tee e eee 40 35 MDSIBDBS noe nepote i ald o iR epe eee edes 41 36 MOVEMENT i euteepesee eR E e i naan 42 3 7 Other output files ecce ec ee et eg ree Re eu eod 42 4 Lhebasiecalculations eere eR HERREN HORE NEM Ee LEO SAO eee CENE Ee sa ed 44 4 1 Self consistent calculations JOB ZSCF essere 44 4 2 None self consist calculations JOBZNONSCE eene 44 4 3 Density of States Calculations JOB DOS sese 46 4 4 Atomic Relaxation JOB RELA X esses nennen 47 4 5 Nudged Elastic Band Calculations JOBZNED sees 48 4 6 Molecular Dynamics JOB MD
25. F the default not calculate the force and output the file 18 IN SYMM IN SYMM T PWmat will use the file IN SSYMM the name is fixed to perform symmetry operations The PWmat supports space group symmetry for crystals IN SYMM should contain space group symmetry operations IN SYMM is usually generated together with IN KPT by running check x which will also check whether the IN SYMM exists if INSSYMM T IN SYMM F PWmat will not use any symmetry operations and not use IN SYMM This is the default value IN KPT IN KPT T PWmat will use the k points from file IN KPT which contains the k points and their weights The IN KPT can be generated together with IN SYMM by running check x with information from variable MP N123 In band structure calculation it can be generated by check x together with KP INV in etotinput It can also be edited manually IN KPT F PWmat will not use the file IN KPT and it will only use Gamma point This is the default value MP N123 MP_N123 nk1 nk2 nk3 sk1 sk2 sk3 This variable is only used by check x not actually used by PWmat This is the Monkhorst Pack grids to generate the reduced k points When running check x If this line is provided the check x will generate the IN SYMM and IN KPT using the above Monkhorst Pack parameters and store them in IN SYMM IN KPT If no MP NI23 and KP INV is provided
26. LGORITHM lines It can terminate the CG steps before the 20 nlineO nlinel have been reached E ERROR The error tolerance convergence criterion for the total energy Hartree in the SCF iterations The default value is 2 10 eV This is related to the SCF ITERO SCF ITERI lines It can terminate the SCF iteration before the maximum steps niterO niterl have been reached RHO ERROR The error tolerance convergence criterion using the SCF iteration difference between the input and output charge density If the relative error of input and output charge density in one SCF step is less than RHO ERROR the SCF iteration will be stopped The default value is 1 5E 6 A variable to control the stopping of the internal CG iterations This is to estimate the charge density error due to the wave function of CG iteration error The estimated charge density error should be less than output input SCF charge density error multiplied by w cg in order to stop the CG steps The default value is 7x102 Smaller this value more stringent requirement as a result more likely this is not used W SCF A variable to control the stopping of the SCF iterations This is to estimate the atomic force error due to the charge density error of SCF iterations The estimated force error should be smaller than the previous MD or RELAX step force multiplied by W SCF in order to stop the SCF iterations Smaller this value
27. NSC 63446274861377E 02 0 2667E 00 sum_i occ i eigen i eV_ E rho V_Hxc 10861401146434E 04 2666E 00 WntV Hxc r rho r dr V Hxc hartree exchange correction eV E Hxc 0 83676461716761E 03 0 2488E 00 The sum of Hartree exchange and correlation energies eV TS 40772013136366E 02 9536E 04 occupation entropy term eV E_tot eV 23461766740827E 04 0 2133E 04 total energy and E_tot this step E_tot last SCF step eV E_tot Ryd 17244074436392E 03 0 1567E 05 total energy and E_tot this step E_tot last SCF step Ryd Zero temp E_tot 23461746354821E 04 Using formula E_tot T TS N 2 N O entropy corrected T 0 energy E_Hart E_xc E_ion 0 1199722E 04 3629580E 03 287406E 04 E Hart Coulomb interaction energy eV E rhoVext E IVext 0 000000E 00 0 00000E 00 E_rhoVext int V ext rho dr E lVext lon V ext energy eV E_psiV E_dDrho 3341520719E 03 6170243465E 0 E dDrho sum_i Dj1j2 lt beta_j1 psi_i gt lt psi_i beta_j2 gt eV ave vtot vO 3533819079E 00 Ivo eV ave V_ion_s or p d ave V_Hatree 0 ave Vtot ave V_xc v0 E_psiV int rho Vtot dr eV mch_pulay drho_in drho_out 0 2559E 04 0 1447E 04 rho_in rho_out before and after mch_pulay xk RESULT atom_move_step E_tot 0 0 234617667408274E 04 finished input atom config calc following are atomic relaxation kk ok K K ok ok K oe ok oe oe ok ok ok K K oe oe ok K oe K K K
28. SCF IN PSPI Si upf The first two numbers are nodel node2 Nodel is the number of processors for plane wave parallelization Nodel should be able to divide n1 n2 n1 n2 n3 are FFT grid in real space Node2 is the number of processor groups for k point parallelization It will divide the number of k points into node2 groups IN ATOM indicates the atomic position file atom config as an PWmat etot input convention all the input file tags will have the form IN XXX while all the output file tags will have the form OUT XXX JOB tells the PWmat what kind of jobs to run IN PSPI indicates the name of pseudopotential file of first atom type the second atom type will be IN PSP2 The etot input has a format of tag name xxx value The orders of different tags can be changed except the first line nodel node2 The names of the variables are case insensitive Line which starts with will be an annotation line After the xxx value one can add annotations in the same line as well e g MP N123 4 4 4 0 0 0 Monkhorst Pack parameter without shift If one runs gt check x a long version of etotinput etotinput ong will be generated This etot input long can be copied as etot input used as the input file This etotinput long gives an explicit form for all the parameters to be used in the calculation generated by defaults An experienced user can change some of the parameters according to the need However one can also just
29. So there are in total Nimage 2 configurations in the string of images connection the initial and final configurations ak the spring constant for the image string eV A In the NEB a string connecting the images are used to ensure the coverage between the initial and final configurations Ak 0 1 to 1 eV A2 are reasonable values Larger ak especially for type spring 2 better the convergence but it can introduce bigger errors for type spring 2 type spring the type of string used in NEB algorithm 1 the original NEB algorithm where the string force perpendicular to the string tangent is removed 2 a conventional string the perpendicular string force is not removed Type spring 2 converges better but it can introduce an error larger ak larger the error But one can first use larger ak then after the initial NEB relaxed re runs NEB using smaller ak or type_spring 1 This will help the convergence E0 EN the precalculated e g using JOB RELAX initial EO and final EN local minima energies in eV for configurations in atom config and atom2 config Actually these numbers are not used in the algorithm but will make plotting more straight forward itype_at2 atom2 config the type of atom2 config file and the atomic position file name atom2 config Itype_at2 1 atom2 config is the second minimum configuration the first 11 local minimum configuration is given in IN ATOMz atom config Then f
30. as the following format 64 Lattice vector 0 1084993850E 02 0 0000000000E 00 0 0000000000E 00 Position 30 0 952534560 30 0 540553000 16 0 242857140 Force 30 0 060040948 30 0 001068674 0 0000000000E 00 0 0000000000E 00 0 1084993850E 02 0 0000000000E 00 0 0000000000E 00 0 1084993850E 02 0 363594470 0 382027650 0 850230410 0 966359450 0 140553000 0 68433 1800 1 1 1 0 097096690 0 063013193 0 002521614 0 000147553 27 16 0 007955164 0 008758074 0 029047748 Velocity 30 0 02339881 0 287387433 0 109339839 30 0 23878474 0 210836551 0 0493 LIT 16 0 53761771 0 023987172 0 288399911 They have the following meanings Natom the number of atoms in the system as a result Position Force Velocity sections will all have Natom lines each atom for one atom Lattice vector The header of the lattice vector AL 3 3 section There will be three lines following Lattice vector AL 1 1 AL 2 1 AL 3 1 the first vector of the super cell edge in Angstrom AL 1 2 AL 2 2 AL 3 2 the second vector of the super cell edge in Angstrom AL 1 3 AL 2 3 AL 3 3 the third vector of the super cell edge in Angstrom Position the header of the atomic positions of the system There will be Natom lines following Position each line describe the position of one atom in the following form Zatom xl x2 x3 imv1 imv2 imv3 30 0 2293 0 59822 0 44444 1 1 1 Zatom is the atomic
31. ave function charge density file They are all binary file They can be copied as IN XXX file as input for the next run or they can be visualized Their 42 formats are the same as their IN XXX file see Section 2 5 Other input files The format for OUT DENS is the same as for OUT RHO The bpsiiofil10000x file are the wave function to atomic orbital projection file for kpoint x These are used to generate the partial DOS It can be removed after DOS is generated 43 4 The basic calculations 4 1 Self consistent calculations JOB SCF This is a one shot DFT calculation for a fixed atomic position It can be used to study the total energy the magnetic moment the charge density the electronic structure etc It will not move the atoms The charge density will be iteratively calculated until it converges input charge density equals the output charge density If subsequent runs e g for NONSCF or DOS are expected one should set OUT WG T OUT RHO T OUT VR T they are all defaults for SCF runs so they will output files OUT WG OUT RHO OUT VR for subsequent uses For band structure plot and DOS plot remember to copy its REPORT to REPORT SCF so the Fermi energy in REPORT can be read out Note in JOB SCF calculation in order to the SCF calculation the icmix for the late SCF iterations must be 1 indicating there will be charge mixing and charge update in the following segment of the etot input file SCF_ITERO
32. ction step Its search direction p is the same as in previous steps all the way to the last NEW step Note the energy of this trial step can be higher than previous step So to see the convergence only the energies for the NEW steps should be used E total energy of this step in eV Av F M F average and maximum atomic forces eV A dE the SCF iteration E n E n 1 eV This is used to judge whether the SCF iteration is converged Note this is not the dE between this relaxation step and previous relaxation step dRho the SCF iteration Irho n rho n 1 I relative error This is used to judge whether the SCF is converged SCF the SCF iteration number for this step dL the movement IR R new initial of this step in atomic unit Bohr R new initial 39 is the initial atomic position of this line minimization direction Note for the NEW step there is already one dL In another word the dL shown in the NEW line is actually the IR R new initial for the R in the following CORR line i e it is the length of the first trial step Similarly the dL of one CORR line is the dL of the R in the following line The dL s in the NEW and subsequent CORR before the next NEW lines are in the same search direction and all measured from the beginning point of this new line direction p F the force of the current step project to the search direction Note the purpose of the line minimization is to make p F zero it uses lin
33. des2 check x can be used to generate IN KPT and IN SYMM from Monkhorst Pack parameters MP NI123 nkl nk2 nk3 skl sk2 sk3 in etot input At the same time one needs to add IN SYMM T IN KPT T in etot input in order to use these k points and symmetry files Check x can also be used to generate IN KPT for band structure calculations when KP_INV M is used here M are the number of K points between two special k points which is determined by check x following the symmetry of the supercell shape 5 Run PWmat e g mpirun np num PWmat or just gt PWmat gt out amp when there is only on GPU in the machine Note num the number of GPU CPU processors must be equal to nodel node2 In our code one CPU is bundled with one GPU 6 Post processing e g gt convert_from_config x lt MOVEMENT or gt convert from config x atom config will change the config or MOVEMENT file to xyz and xsf file for visualization One can also run convert rho x OUT RHO note NO lt between convert rho x and OUT RHO to generate RHO xsf for isosurface plot of the charge density 7 Visualize the structure or make an animation with the xyz and xsf file using VMD to watch the animation and VEST to view one configuration For Mstation users all the above programs are pre installed in the Mstation As discussed above besides the main code PWmat we also provide many utility codes check x convert t
34. dopotential format Note that due to slightly different implementation of the nonlocal pseudopotentials the PWSCF and PWmat energies and forces might be slightly different But the difference is rather small With the compatibility of PWmat and PWSCF one can fully take the advantages of the wide functionalities of the PWSCF while enjoy the speed of PWmat for some of the key calculations The available PWSCF capabilities include Wannier function 50 generation linear response phonon band structure calculation linear response TDDFT calculation GW calculation We refer the user to consult the open source PWSCF manual for how to calculate these properties To generate the PWSCF compatible wave function and potential files set PWSCF T in etot input 51 5 Pre and post processing programs 5 1 Pre processing We provide the utility programs to help the preparing of the PW mat input files 5 1 1 convert to config x One useful program is convert to config x which convert the atomic position file from other formats cell xsf vasp to config file format which we use We also provide a tutorial for how to generate cell xsf or vasp file from visualization tools or online database 5 1 2 check x Another major preprocessing tool is check x it will check the etot input file to see whether there is any error in it e g not missing any line it will generate a etot input long spelling out all the defaul
35. e atomic force as reported in RELAXSTEPS one can use type string 2 otherwise just use type string 1 Place the EO EN from previous RELAX calculations to the line NEB DETAIL Use itype at2 1 Then do the JOB NEB calculation If type_string 2 was used and ak is a bite large one can do another calculation 3 copy MOVEMENT into atom continue remove all the previous iterations Replace 1 atom2 config in NEB DETAIL by 2 atom continue Now either use type string 1 true NEB method or still use type string 2 but with a much smaller ak Do the calculation again Check NEB BARRIER make sure the images are roughly in equal distance and the angle between R image 1 R image and R image R image 1 especially around the saddle point is close to 0 degree costheta close to 1 4 6 Molecular Dynamics JOB MD This is for ab initio molecular dynamics simulations There are three methods md 1 Verlet algorithm for energy conserved true Newton dynamics md 2 Langevin dynamics with a viscosity and a thermos bath for given temperature control simulation md 3 Nose Hoover stochastic methods for given temperature control simulation We recommend to use either md 1 or md 3 This simulation will output MDSTEPS and MOVEMENT One can continue the simulation by copying MOVEMENT to atom config remove all the other iterations except the last one then restart the calculation only retype the running commands in the terminal
36. e the G space sphere and nl n2 n3 grid Nodel must evenly divides n1 n2 which will be provided by check x Node2 The number of processor groups to divide the k points The best is to have node2 evenly or almost evenly divide the number of k points Note the total number of processors must equal to nodel node2 IN ATOM IN ATOM atom config The atomic positions file Its specification is described in the section atom config of this manual IN PSP1 2 The names of the pseudopotential files Such as IN PSP1 Na ncpp upf for norm conserving IN PSP1 Na uspp upf for ultra soft Currently different types of pseudopotentials cannot be mixed i e either all atoms are in norm conserving PSP or all in ultra soft The PWmat can use norm conserving or ultrasoft pseudopotentials We only read upf formatted pseudopotential files UPF format is the Quantum Espresso format We provide a tested package of pseudopotentials One can also copy pseudopotential files from the Quantum Espresso website and use it as it is JOB Controls what PW mat will do JOB SCF do only the self consistent field iterations self consistently determine the charge density will output the total energy It will not move atoms JOB NONSCF do non self consistent calculations lt usually inputs a converged potential or charge density only calculates the eigen wave functions non self consistently It
37. e x1 x2 x3 value 0 1 are the fractional coordination values in the unit cell edge vectors 1 2 3 used to cut a box for this special Coulomb solution In the other word the center of the box is at x 140 5 x2 0 5 x3 0 5 COULOMB 11 x1 A slab calculation along the first direction with the cut at xl This can avoid the image interaction between slabs COULOMB 12 x2 A slab calculation along the second direction with the cut at x2 COULOMB 13 x3 A slab calculation along the third direction with the cut at x3 OUT WG OUT WG T TRUE PWmat will output a file OUT WG which stores the final wave functions in G space When SPIN 2 an extra file OUT WG 2 will also be output This is the default value 15 OUT WG F FALSE will not output the wave function file IN WG IN WG T PWmat will read in the initial wave functions in G space from the file IN WG e g from previous calculation When SPIN 2 an extra file IN WG 2 will also be read in IN WG F the default the PWmat will start from random wave function OUT RHO OUT RHO T PWmat will output a file OUT RHO the final charge density in real space grid n1L n2L n3L This is the default value If SPIN 2 PWmat will write out an extra file OUT RHO 2 OUT RHO F not output the charge file IN RHO IN RHO T PWmat will read in the initial charge density from fi
38. ear interpolation of p F to predict the next step size dL in this line minimization p FO the same as p FO however not use the force of this configuration but use the force of R new initial Thus this p FO is the same throughout one line minimization direction Fch the force check calculated as dL F F0 2 dE dL is the displacement for this step from the R new initial F FO are the forces are the two ends of this step F is the force at the current position FO is the initial force at R new initial dE is the total energy difference of this step the current energy minus the initial energy at the beginning of the new search direction Fch 1 indicates all the calculations are accurate Note for single precision GPU calculation the dE is usually less accurate than dL F F0 2 so Fch not equaling 1 can still be fine since the force is good and the relaxation algorithms are based on force not the total energy 3 4 NEB BARRIER The NEB BARRIER is the file concisely report the energies along the images for different relaxation iteration steps One can yield the barrier height and profiles from NEB BARRIER It has the following format iter 19 Etot eV dist Bohr angle cos th 0 0 75306186045042E 03 0 504486E 00 0 000000E 00 1 0 75305820517778E 03 0 520270E 00 0 944578E 00 40 2 0 75304052843358E 03 0 530724E 00 0 846617E 00 3 0 75296036069356E 03 0 526520E 00 0 355627E 00 4 0 75266754347227E 03
39. elf consistent step The default value of nlineO is 4 Note the CG line minimization can be stopped if the error is smaller than WG_ERROR or the condition specified by w_cg is reached So the stopping of CG iterations is controlled by three parameters nlineO WG_ERROR w_cg whichever is satisfied first Algorithm0 Algorithm0 imth icmix dE Fermi Dirac imth icmix dE Fermi Dirac imth icmix dE Fermi Dirac These are the niterO lines immediately following the SCF ITERO line It describes the detail procedures for each SCF step imth 1 the old band by band CG algorithm It should not be used unless for some special situation imth 3 the all band conjugate gradient method This is the default method imth 2 the DIIS method This could be faster than imth 3 but could also have stability problems It should only be used in SCF iteration steps where the wave function is in some degree converged e g not for random wave functions icmix 0 no charge mixing and update at this SCF step In other word at this step it is a NONSCF step For JOB SCF RELAX MD by default for the first three SCF steps icmix 23 0 and icmix 1 for subsequent steps For JOB NONSCF for all steps icmix 0 icmix 1 with charge mixing and update for this SCF step dE the kT equivalent energy in eV for Fermi Dirac formula to calculate the electron occupations of the eigen wave functions accord
40. ing to their eigen energies si The default value is 0 1eV Fermi Dirac with possible values 1 2 3 4 5 Different formulas for the Fermi Dirac equivalent function to calculate the wave function occupation using i and dE These formulas are 1 Fermi Dirac 2 Gaussian 3 4 5 Gaussian with other prefactor polynomials The default value is 1 SCF_ITER1 SCF_ITER1 niter1 nlinel This is for subsequent SCF calculations for JOB RELAX MD except the first SCF step It has the same meaning as in SCF ITERO The default values are niter1 6 nline1 4 Algorithm Algorithm1 imth icmix dE Fermi Dirac imth icmix dE Fermi Dirac imth icmix dE Fermi Dirac These are the niterl lines immediately following the SCF ITERI line It has the same meaning as the algorithm0 Usually however the icmix is always 1 For default imth 3 icmix 1 dE 0 1 Fermi Dirac 1 24 NONLOCAL The nonlocal is the nonlocal pseudopotential implementation flag nonlocal 1 no nonlocal potential nonlocal 2 the default real space nonlocal pseudo potential implementation It used the mask function method nonlocal 3 G space nonlocal pseudo potential implementation Rcut The Rcut in Bohr unit note 1 Bohr 0 529177 10 m is for the cut off radius for nonlocal pseudopotential implementations It defines the core radius of the nonlocal part The default value is 3 2 for all calculations except
41. irst line is the explanation of this symmetry operation The following three line defines the point group rotation matrix s 3 3 around the origin x1 x2 x3 0 0 0 point The rotation s 3 3 will convert a real space point x1 x2 x3 in lattice cell fractional coordination into another point following yl s d 1 x1 s Q 1 x24 s 3 1 x3 y2 s 1 2 x1 s 2 2 x2 s 3 2 x3 y3 s 1 3 x1 s 2 3 x2 s 3 3 x3 The next line defines the fractional translation in the space group Thus we have yl yl 1 y2 y2 1 2 y3 y3 1 3 2 6 Other input files 33 One might use other input files e g IN WG wave function input file IN RHO charge density input file IN VR potential input file IN VEXT external potential file These are usually generated from the previous SCF calculations e g can be copied over from the corresponding OUT XXX files They are binary files The internal format for wave function file IN WG is complicated it is the wave functions in G space we usually do not view it only copy it from OUT WG to IN WG for next run For IN RHO IN VR IN VEXT internally it has the following format can be written or read out like this Write in xx n1L n2L n3L nodel Write in xx AL Do iproc 1 nodel Write in xx vr ir nr nL iproc 1 ir 1 nr nL Here nr n nlL n2L n3L nodel Vr ii is a 1D array form of vr i j K with ii to i
42. j K correspondence as ii i 1 n2L n3L j 1 n3L k 34 3 Output files 3 1 Standard output file Standard on screen output contains the verbose information of each SCF calculation Standard output has almost all the information for PWmat but it might be messy to read If we use gt mpprun np num PWmat gt out amp to run our job the standard output will be stored in the file out 3 2REPORT The REPORT file contains the most useful information in a concise way for the run It might have the following format with explanation given at the right hand side in blue 5 1 IN ATOM 064 config JOB RELAX RELAX DETAIL 1 100 0 367E 03 IN PSP1 uspp Zn txt IN PSP2 uspp S txt Ecut 21 0000000000000 Ecut2 42 0000000000000 Ecut2L 168 000000000000 N123 45 45 45 NS123 45 45 45 N123L 90 90 90 SPIN 1 XCFUNTIONA 0 000000000000000E 000 SMTH 1 00000000000000 COULOMB 0 PWSCF F 35 IN WG F OUT WG F IN RHO F OUT RHO F IN VR F OUT VR F IN VEXT F OUT RHO_SP 0 OUT FORCE T IN SYMM F IN KPT F 576 000000000000 NUM Electron NUM BAND 312 WG ERRO 1 000000000000000E 004 E Error 5 760000000000000E 005 w cg 7 000000000000001E 002 W SCF 1 000000000000000E 003 QijLO GS 1 SBFFY 1 SCF_ITERO 20 4 ALGORITHMO 3 0 0000 0 10000 1 3 0 0000 0 10000 1 3 0 0000 0 10000 1 3 1 0000 0 10000 1 3 1 0000 0 10000 1 3 1 0000
43. le IN RHO stored in the real space grid n1L n2L n3L Note if both IN VR and IN RHO are set to T the program will use the read in potential to start the calculation If SPIN 2 PWmat will read an extra file IN RHO 2 IN RHO F not input the charge density This is the default OUT VR OUT VR T PWmat will output the total potential in file OUT VR stored in real space grid n1L n2L n3L This is the default value When SPIN 2 an extra file DUT VR 2 will be output 16 OUT VR F not output the potential file IN VR IN VR T PWmat will read in the initial potential from file IN VR in real space grid nIL n2L n3L Note if both IN VR and IN RHO are set to T the program will use the read in potential to start the calculation When SPIN 2 an extra file IN VR 2 will also be read IN VR F not read in the file This is the default IN VEXT IN VEXT T PWmat will read in an external potential from file IN VEXT in real space grid n1L n2L n3L Both the total energy and forces are calculated using this external potential IN VEXT F no external potential is used This is the default value OUT RHO SP OUT RHO SP iflag kpt1 kpt2 ispin1 ispin2 iwl iw2 Controls the output of partial charge density or wave function without square in real space grid nl n2 n3 from selected eigen orbitals within the intervals k points
44. more accurate SCF is used The default value for JOB RELAX is 0 003 the default value for JOB MD is 0 02 21 SCF ITERO SCF_ITER0 niter0 nlineO These variables control the charge density self consistent iterations for the first SCF run for JOB SCF RELAX MD For RELAX MD the first step SCF run with the initial atomic positions uses SCF_ITERO and subsequent steps for moved atomic positions uses the values of SCF ITERI They are set differently because normally the first run requires much more steps It is also used for NONSCF run in which the icmix 0 for all the niterO lines in the following This variable is not used for JOB DOS niter0 the number of self consistent SCF iterations steps There must be niterO lines in etot input following this line in the form of ALGORITHMO imth icmix dE Fermi Dirac imth icmix dE Fermi Dirac imth icmix dE Fermi Dirac See ALGORITHMO for explanation The Default value for niterO is 20 Note the SCF iteration can be stopped before the niterO has been reached if the E Error has been satisfied or the condition specified by w scf has been reached So the stopping of SCF iteration is controlled by four parameters niter E error RHO error w scf whichever is satisfied first nline0 the number of CG line minimization steps to solve the 22 wave functions according to Hy y for a given potential hence H at each charge s
45. nts between the two end points atom1 config atom2 config This can guarantee the path can go from one configuration to the other configuration To avoid the force from the potential DFT energy to move the images away from the barrier saddle point which lower the potential energy the tangent component of the potential force will be removed To avoid the force of the elastic string from moving the string away from the saddle point corner cutting since an elastic string like to have the minimum length the perpendicular to the string tangent component of the string force will be removed So the force of the string will only maintain equal distances between the images This is the essence of the NEB However after such modification there is no guarantee the remaining total force can be written down as a gradient of a potential i e the vortices of the force might not be zero This can cause significant difficulty in the relaxation procedure to make the force zero e g if one just follows the force to make atomic movement it is possible that the relaxation iteration can end up in an infinite loop Another common problem is that string is not smooth with the vector R image l R image and R image R image 1 having an angle not close to 180 degree Thus there are several points need to be considered when doing a JOB NEB calculation First it is better to use imth 3 steepest decent for the atomic relaxation method This is because the CG
46. o config x convert from config x etc In the online page http www pwmat com pwmat tutorial we provide detail tutorials for how to use a few open source visualization software packages both for preparing the input files pre processing and analyzing and viewing the results post processing Note due to rapid development there could be some minor changes for the tutorial e g the pseudopotential file should be atom upf instead atom vwr etc But the basic steps and ideas are the same and the changes should be obvious Thus we provide an easy to follow graphically based workflow from the beginning to the end for doing actually research using PWmat Although these visualization tools are open sourced we provide the utilities to link them together and to make them work together seamlessly Please check the workflow in Appendix A 2 Input files PWmat needs a few basic input files to start the calculation atom config etot input and atom upf pseudopotential files While the pseudopotential files can be copied from the library for the required atoms atom config and etot input have to be prepared specifically for the calculation In the following we explain the etotinput and atom config files 2 letot input The etot input is the most essential file controlling how to run PWmat and what are the names of the other input files Here is an example of a short essential etot input file 2442 IN ATOM atom config JOB
47. or BFGS method which assumes a parabolic potential and the related force might no longer work in NEB Second we have implemented two types of string to deal with the string force type_string l1 means the original NEB string remove the perpendicular string force while type_string 2 means the conventional string the perpendicular string force is not removed If the problem failed to converge one possibility is to first use type_string 2 and a relative large string constant ak Then after that is converged one can do a second NEB calculation copy MOVEMENT to atom2 config and use itype_at2 2 using either itype_string 1 or a smaller ak while still use type_string 2 All these choices are to increase the flexibility in NEB calculations A normal NEB calculation should have the following steps 1 using JOB RELAX to calculate the two local minimum their atomic positions atoml config atom2 config and energies El E2 Note the atomic orders in atoml config and atom2 config must be the same 2 Use JOB NEB and write NEB DETAIL as IN ATOM atom1 config NEB_DETAIL imth nstep force_tol Nimage ak type_string E0 EN itype_at2 48 atom2 config Make sure imth 3 nstep force tol can be similar as in RELAX DETAIL Nimage is the number of images between the first and last images Typically Nimage can be 5 to 10 Choose a string constant Typically ak 0 1 to 1 eV A sounds reasonable If the relaxation is difficult to reduce th
48. rom atom config to atom2 config Nimage equal distance images will be created by linear interpolations Itype_at2 2 atom2 config contains all the Nimage 2 image configurations most likely from a previous unconverged NEB run and copied from MOVEMENT including the initial and final images Thus Itype_at2 2 is a continued NEB run following the previous NEB runs In this case the atom config in IN ATOMczatom config is not used but that line still need to be provided Note in this case the atomic positions files for each image inside atom2 config must have first the Position section followed by Force section even though atomic forces are not used That is the format output in MOVEMENT under JOB NEB See NEB BARRIER for JOB NEB output file NSCALE_VVMD NSCALE_VVMD NSTEP To scale the kinetic energy in Verlet_MD every NSTEP steps so the total energy is conserved The default NSTEP is 100 RELAX_DETAIL RELAX_DETAIL imth nstep force_tol This is an optional line for JOB RELAX It controls the atomic relaxation steps Note in the current atomic relaxation the unit cell will be kept fixed Imth 1 or 2 or 3 indicate the method of relaxation 1 for conjugated gradient 2 for BFGS method 3 for steepest decent this is mostly for JOB NEB The default is imthz1 Nstep is the number of relaxation steps each total energy calculation is one step i e 12 it counts the steps inside
49. s and connect them into a band structure path The resulting k points will be written in IN KPT This can be used in the JOB NONSCF to generate the band structure 5 2 Post processing The PWmat package also provides post processing programs to facilitate the post processing of the runs Mainly it has the following post processing programs 5 2 1 convert from config x This program will convert the xatom config file or MOVEMENT into xyz and xsf formats so they can be used by different visualization tools for viewing We provide a 53 tutorial for how to use these visualization tools This is the sister program of convert to config x in the pre processing 5 2 2 plot band structure x Before run the plot band structure x please prepare OUT EIGEN IN KPT and REPORT SCF this file is copied the SCF calculation output REPORT into REPORT SCF Then it will generate the following files band eps band jpb and origin dat the data file of band structure which can be used to plot band with specified scale and regions 5 2 3 plot dos x This program uses the DOS totalspin with 0 1 eV energy broadening If one want to shift the Fermi energy to zero the REPORT SCF is needed After running the plot dos x it will generate dos eps or dos jpg However if one wishes to change this broadening or more importantly plot the atomic selected partial DOS one can use this plot DOS new x To do that one needs to provide a modified
50. t see below A concise output is reported in MDSTEPS The atomic movements for every step are reported in MOVEMENT MD DETAIL MD DETAIL MD Mstep dt Temp1 Temp2 Note this is a required line for JOBZMD There is no default values hence must be input by hand MD the method of MD algorithm 1 Verlet 2 Langevin 3 Nose Hoover Verlet is for NVE and Langevin and Nose Hoover are for NVT Currently we do not have NPE or NPT MDs Mstep the number of MD steps dt the time length for each MD step in the unit of fs Ifs 1 10 s Note usually with H atoms dt should be 1fs and with heavier atoms dt could be 2fs Templ the beginning temperature in K Temp2 the final temperature in K Temp2 is only for MD 2 3 During the MD the program will adjust the temperature let it goes from Templ to Temp2 For Verlet Temp2 is not used NEB DETAILS NEB DETAIL imth nstep force tol Nimage ak type string E0 EN itype at2 atom2 config imth 1 2 3 the algorithm used for atomic relaxation 1 conjugated gradient 2 BFGS 3 deepest decent For NEB calculation it is better to use 3 otherwise it might not converge nstep the maximum number of line minimization steps in the relaxation process 10 force tol the force tolerance eV A to stop the relaxation Nimage the number of images in the NEB method these are the images except the initial and final two valleys
51. t parameters to be used in the calculation If wished one can copy etot input long into etotinput manually change some parameters One can also use the original etot input for PWmat runs Although not necessary we do strongly encourage the user to use check x for the first time to calculate one system One important thing is that check x will spell out the FFT grid n1 n2 n3 That will help the user to choose node first line in etot input which must evenly divide n1 n2 So one might want to change nodel in etot input according to n1 n2 n3 or change n1 n2 n3 according to nodel 52 Another thing one has to use check x is to generate IN KPT IN SYMM if one wants to use k points and symmetry The Monkhorst Pack parameters MP N123 nkl nk2 nk3 sk1 sk2 sk3 or alternatively the band structure parameter KP INV2m must be included in etot input in order for check x to generate IN KPT IN SYMM IN KPT T IN SYMM T must be used in order to use these k points and symmetry file Thus in other words in order to use k points and symmetry point one must run gt check x first Note IN KPT and IN SYMM should be used together Another useful function of check x is to generates the IN KPT for band structure calculations After setting the KP INV in etot input it will take the information of the atom config file match that to one of the 14 primary cell shapes then generate the corresponding high symmetry k point
52. then when run check x no IN KPT and 19 IN SYMM will be generated Thus if don t want to generate k points but want to generate symmetry one needs to set MP_N123 111000 The skl sk2 and sk3 must be either 0 no offset or 1 grid displaced by half a grid point in the corresponding direction KP_INV The number of interval points between two high symmetry k points It conflict with MP _N123 i e the user have to use only one of them when necessary This is for band structure calculation When check x is run it will use the information of the unit cell edges generate the special k points and link them with lines with KP_INV points Note KP_INV will also trigger check x to generate IN KPT and IN SYMM NUM_ELECTRON The total number of occupied valence electron in the system One can use this to make the system charged or not charged Note for charged system calculations a uniformed back ground charge is used to solve the Possion equation for COULOMB 0 Default value is the value for neutral system NUM_BAND The number of orbitals to be calculated When SPIN 2 there are NUM BAND spin up orbitals and NUM BAND spin down orbitals The default value is about min 1 2 Num_electron 2 Num_electron 2 20 WG_ERROR The error tolerance convergence criterion for the wave function conjugate gradient iterations Hartree The default value is 10 This is related to ALGORITHMO and A
53. ulated ug ilocal 2 rcut 3 20 ntype 2 lilocal pseudopot implement rcut core cutoff a u ntype psp Type numkpt 1 num sym 1 numkpt num of k points num sym num of symmetry operations KKK K K K K K K K K K K K K K K K K K K AL1 AL2 AL3 in x y z 10 8499385 0 0000000 0 0000000 in amstrong 0 0000000 10 8499385 0 0000000 0 0000000 0 0000000 10 8499385 FKK K K ok ok ok K oe K oe oe K ok K K K K K K K K K K iter 4 ave_lin 4 0 iCGmth 3 liter SCF iter num ave line CG line min num Ef eV 0 3693964E 01 Fermi energy err of ug 0 9756E 03 Avery wave function ug error H e ug dv_ave drho_tot 0 0000E 00 0 6641E 01 E_tot 23469213955264E 04 2347E 04 iter 10 ave_lin 2 0 iCGmth 3 iCGmth the method for wave func Solver 3 CG 2 DIIS Ef eV 0 3896764E 01 err of ug 0 1108E 03 dv_ave drho_tot 0 7602E 03 0 4362E 03 V in V out and rho_in rho_out errors in SCF a u E tot 23461766740827E 04 0 2133E 04 Total energy in eV E thisstep E laststep eV E Fermi eV 3 89676368689153 Ef eV 0 3896764E 01 dvE dvE n dvE n 1 0 5779E 06 0 3464E 06 dvE Vint V_in V_out rho r dr a u dv ave drho tot 0 7602E 03 0 4362E 03 err of ug 0 1108E 03 Ewald 12541186504745E 04 Ewald energy eV Alpha 0 57532443097826E 02 Pseudopotential Alpha energy eV E_extV 0 00000000000000E 00 0 0000E 00 energy due to ext potential int V_ext r rho r dr eV 37 E
54. use the short version etot input to run PWmat using the default values for most parameters Note when running PWmat it will generate file REPORT The first part of the REPORT file is the same as the etot input long So one can also copy this REPORT to etot input delete the others lines use it as an explicit input file One example of long etot input file looks like this 4 2 nodel node2 IN ATOM atom config JOB SCF IN PSP1 Al upf Ecut 20 0000000000000 Ecut2 40 0000000000000 Ecut2L 40 0000000000000 NI23 16 16 16 NS123 16 16 16 NI23L 16 16 16 SPIN 1 XCFUNTIONAL 0 000000000000000E 000 Num Electron 12 0000000000000 Num_band 16 WG_ERROR 1 000000000000000E 004 E Error W cg w_scf SCF_ITERO ALGORITHMO SCF_ITERI ALGORITHMI NONLOCAL Rcut 8 000000000000000E 007 7 00000000000000 1 E 002 0 000000000000000E 00 20 4 3 0 0000 0 10000 3 0 0000 0 10000 3 0 0000 0 10000 3 0 0000 0 10000 3 1 0000 0 10000 3 1 0000 0 10000 3 1 0000 0 10000 6 4 3 1 0000 0 10000 3 1 0000 0 10000 3 1 0000 0 10000 3 1 0000 0 10000 3 1 0000 0 10000 3 1 0000 0 10000 2 3 20000004768372 aS o t t M toto to t t In the following we explain the meaning of each variable in the long version of etotinput The Variable in red are the mandatory variables and all the others are optional will be set automatically by default values Variable Specification Nodel The number of processors to divid
55. xatom input file with the following format 30 0 952534560 0 363594470 0 382027650 1 1 1 wl 30 0 540553000 0 850230410 0 966359450 I 11 w2 16 0 242857140 0 140553000 0 68433 1800 1 1 1 w3 Here wl w2 w3 are the weight for this atom in the partial DOS 54 5 2 4 Convert rho x This program convert the potential or charge density file OUT VR OUT RHO OUT RHO SP to a rho xsf format so they can be view by VESTA To use it type convert rho x OUT RHO It will generate RHO xsf Note in above do not use gt convert_rho x lt OUT RHO 55 Appendix A WORKFLOW xyz and xsf format file i convert to config x i Pre process ees etot input i check x MD RELAX SCF NONSCF DOS plot_band x plot_dos_new x Post process e convert rho x convert from config x Post PWscf calculation 56 Appendix B Useful Websites 1 PEtot http cmsn Ibl gov htmI PEtot PEtot html 2 Quantum Espresso http www quantum espresso org 3 American Mineralogist Crystal Structure Database http rruff geo arizona edu AMS amcsd php 4 NIST Chemistry WebBook http webbook nist gov chemistry 5 XCyrSDen http www xcrysden org 6 VMD http www ks uiuc edu Research vmd 7 Jmol http jmol sourceforge net 8 Nvidia Cuda Zone https
Download Pdf Manuals
Related Search