24小时热门版块排行榜

南方科技大学公共卫生及应急管理学院2025级博士研究生招生报考通知

返回列表

xun枭水草

新虫 (初入文坛)

应助: 0 (幼儿园)
金币: 226.4
散金: 10
帖子: 41
在线: 8.1小时
虫号: 23496637
注册: 2020-09-08
专业: 凝聚态物性 II ：电子结构

[交流] [VASP] 求助：DFPT法算出来的有LO-TO劈裂的声子谱在G点断开了，要如何解决呢已有2人参与

如题，希望能有声子计算方面的高人帮我解决一下这个问题，我真的技穷了
压缩包里的SPOSCAR是3*3*3扩胞的结果，是我计算所使用的。在phonopy进行后处理的时候额外考虑了LO-TO劈裂
INCAR：

NELM=100
EDIFF = 1e-9 ! It is a recommended value. When calculating phonon, 1e-9 at least.
AMIX=0.7
BMIX=0.01

#NBANDS = 96 ! Default = NELECT/2 + NIONS/2
#NPAR=8
#NCORE=6
KPAR=10    !    When irreducible k-points are large in amount, may increase KPAR, which will increase memory usage in most instances and sharply increase speed.
###Parallel settings are related to convergence, but have no certain relation. If electron step fail, change the above three may solve problem.

ISTART = 0
ICHARG = 2 ! For band structures,with ISTART=0 and ICHARG=11. Copying WAVECAR and using ISTRAT=1 may lead to fault in high excited states and lower speed
#LORBIT = 11
#LAMXMIX=4 ! system includes d-element, 4 ; system includes f-element, 6
LWAVE = .FALSE.
LCHARG = .FALSE. ! Sometimes it should be commented
#NWRITE=3

#LREAL=A ! When calculating phonon, do not change it. Default setting is .FALSE.
#ISYM = -1
SYMPREC = 1e-6  ! Default 1e-5
#GGA = PE
GGA_COMPAT = .FALSE. ! Apply spherical cutoff on gradient field.

#ALGO = Fast    !    When precision is high in many aspects, usually need to comment it. NEVER use Exact except some kinds of GW. Exact will decrease speed by thousands of times. When electron step goes wrong, see wiki or comment it to solve.
ISMEAR = 0    ! Gaussian smearing usually, 1 for metal, -5 for DOS or accurate Etot
SIGMA = 0.01    ! Default=0.2, for metals; other should be smaller. For phonon calculation, 0.03 at least
PREC = Accurate  ! Use it in phonon calculation
ENCUT = 450    ! Please set it manually
KSPACING=0.15
ADDGRID=.TRUE.

#LSORBIT = .TRUE.
#LNONCOLLINEAR = .TRUE.

#ISPIN = 2       ! Default=1; 2 for spin-polarized calculations
#MAGMOM = 0 0 0  0 0 0  0 0 0  0 0 0  0 0 0  0 0 0  0 0 0  0 0 0  0 0 0
#LDAU = .TRUE.
#LDAUU = [real array]    ! Coulomb interaction parameters
#LDAUJ = [real array]    ! Exchange interaction parameters

#LHFCALC = .TRUE.    ! Hartree-Fock Calculation. Recommended all bands simultaneous algorithm
#HFSCREEN = [real]

#SCISSOR=
#LVEL = .TRUE.
#LOPTICS = .TRUE.
CSHIFT=0.05
NEDOS=4000    !    Number of grid points for the electronic density of states and dielectric function
#OMEGAMAX=70    !    It can not change the frequency range of calculation for ionic dielectricfunction whatever IBRION=6 or 7 or 8, but can change range for electron dielectricfunction and cause little(or nothing) difference to dielectricfunction's values when the change of OMEGAMAX is not large, so in this case, OMEGAMAX can be used.
   #For ionic dielectricfunction, it is possible for VASP that whole range(highest frequency) can be calculated will be calculated, so to change range, should change accuracy to include higher frequency phonon modes, which means change supercell size. K density, ISYM, NBAND, parallel setting have very little effects on upper limit
#OMEGATL    !    do not effect electron and ion dielectricfunction in every aspect, should not use it.
#DIPOL = [real array] ! center for cell in direct coordinates
#LVTOT = .TRUE.       ! Total local potential
#LVHAR = .TRUE.       ! Containing entire local potential
#LELF = .TRUE.       ! Electron localization function
#LBERRY = .TRUE.       !
#LCALCPOL = .TRUE.
#LCALCEPS = .TRUE.
#EFIELD_PEAD = 0 0 0 ! unit eV/Ang
#LASPH=.TRUE.    !    For Raman's latter half series calculations, use it and LEPSILON.

IBRION=8
LEPSILON=.TRUE.    !    Ionic dielectric function. DFPT need no more empty bands
#use supercell that: 1three lattice constants have little difference, 2lattice constants > 10A, 3atom number > 90. Every result relying on IBRION=5-8 must use supercell.
#When ISYM=2, do not support NCORE!=1, but support kpoints parallel. Kpoints parallel do not change dielectricfunction's values. When ISYM=-1, support band parallel. This role also applies to Raman.
#LPHON_POLAR=.TRUE.    !    Static dielectric tensor and Born charges that are needed for LO-TO splitting are also need supercell calculation.
#PHON_DIELECTRIC=

#PHON_BORN_CHARGES=

回复此楼