计算化学公社
标题: 自己测试例子cp2k与VASP运行速度对比发现cp2k计算时间更长怎么修改cp2k输入文件 [打印本页]
作者Author: 嘿咻嘿咻跑run 时间: 2023-7-17 16:56
标题: 自己测试例子cp2k与VASP运行速度对比发现cp2k计算时间更长怎么修改cp2k输入文件
问题现象描述：
各位好，我最近自己分别用cp2k和VASP跑了一个H在SiC四面体间隙中的间隙形成能的例子，想要测试一下cp2k对比VASP的计算速度，两个计算都考虑了自旋，且K点为4*4*4，晶胞加上H一共有65个原子，cp2k采用的是DZVP-MOLOPT-SR-GTH基组以及GTH-PBE赝势，cp2k输入文件通过Multwfn直接产生，几乎没怎么修改（初步结构优化，后续还用TZVP-MOLOPT-GTH和TZV2P-MOLOPT-GTH基组做了能量计算，能量和VASP对得上），然而目前这一步cp2k耗时1069秒，VASP只用了377秒，cp2k采用cp2k.popt提交，两个计算都采用了36核提交
下面是结构以及cp2k和VASP的主要的输入文件内容
问题：
1.现在cp2k用时比VASP用时高，是由于cp2k对k点的支持问题吗，还是cp2k的in文件有需要改进的地方
2.我现在采取的用DZVP-MOLOPT-SR-GTH结构优化之后，再用TZVP-MOLOPT-GTH进行能量计算的流程是否是合理的
输入文件(cp2k和VASP)
cp2k:（结构部分省略）
&KIND Si
ELEMENT Si
BASIS_SET DZVP-MOLOPT-SR-GTH-q4
POTENTIAL GTH-PBE
&END KIND
&KIND C
ELEMENT C
BASIS_SET DZVP-MOLOPT-SR-GTH-q4
POTENTIAL GTH-PBE
&END KIND
&KIND H
ELEMENT H
BASIS_SET DZVP-MOLOPT-SR-GTH-q1
POTENTIAL GTH-PBE
&END KIND
&END SUBSYS
&DFT
BASIS_SET_FILE_NAME BASIS_MOLOPT
POTENTIAL_FILE_NAME POTENTIAL
# WFN_RESTART_FILE_NAME SiC_222_HTsi-RESTART.wfn
CHARGE 0 #Net charge
MULTIPLICITY 2 #Spinmultiplicity
UKS
&KPOINTS
SCHEME MONKHORST-PACK 4 4 4
&END KPOINTS
&QS
EPS_DEFAULT 1.0E-12 #Set all EPS_xxx to values such that the energy willbe correct up to this value
#EXTRAPOLATION USE_PREV_P #Use converged density matrix of last geometryas initial guess
&END QS
&POISSON
PERIODIC XYZ #Direction(s) of PBC for calculating electrostatics
PSOLVER PERIODIC #The way to solve Poisson equation
&END POISSON
&XC
&XC_FUNCTIONAL PBE
&END XC_FUNCTIONAL
&END XC
&MGRID
CUTOFF 400
REL_CUTOFF 55
&END MGRID
&SCF
MAX_SCF 128
EPS_SCF 1.0E-06 #Convergence threshold of density matrix of inner SCF
# SCF_GUESS RESTART #Use wavefunction from WFN_RESTART_FILE_NAME file asinitial guess
&DIAGONALIZATION
ALGORITHM STANDARD #Algorithm for diagonalization. DAVIDSON is fasterfor large systems
&END DIAGONALIZATION
&MIXING #How to mix old and new density matrices
METHOD BROYDEN_MIXING #PULAY_MIXING is also a good alternative
ALPHA 0.4 #Default. Mixing 40% of new density matrix with the old one
NBROYDEN 8 #Default is 4. Number of previous steps stored for the actualmixing scheme
&END MIXING
&PRINT
&RESTART #Note: Use "&RESTART OFF" can preventgenerating .wfn file
BACKUP_COPIES 0 #Maximum number of backup copies of wfn file. 0 meansnever
&END RESTART
&END PRINT
&END SCF
&END DFT
&END FORCE_EVAL
&MOTION
&GEO_OPT
TYPE MINIMIZATION #Search for minimum
KEEP_SPACE_GROUP F #If T, then space group will be detected andpreserved
OPTIMIZER BFGS #Can also be CG (more robust for difficult cases) orLBFGS
&BFGS
TRUST_RADIUS 0.2 #Trust radius (maximum stepsize) in Angstrom
# RESTART_HESSIAN T #If read initial Hessian, uncomment this line andspecify the file in the next line
# RESTART_FILE_NAME to_be_specified
&END BFGS
MAX_ITER 500 #Maximum number of geometry optimization
MAX_DR 3E-3 #Maximum geometry change
RMS_DR 1.5E-3 #RMS geometry change
MAX_FORCE 4.5E-4 #Maximum force
RMS_FORCE 3E-4 #RMS force
&END GEO_OPT
&PRINT
&TRAJECTORY
FORMAT xyz
&END TRAJECTORY
&RESTART
BACKUP_COPIES 0 #Maximum number of backing up restart file, 0 means never
&END RESTART
&END PRINT
&END MOTION
VASP：
INCAR：
SYSTEM = graphite cell
ISTART = 0
ICHARG = 2
IBRION = 2
ISIF = 2
ENCUT = 450
NELMIN = 2
NELM = 100
EDIFF = 1e-5
EDIFFG = -0.02
NSW = 100
ISMEAR = 0
SIGMA = 0.05
ISPIN = 2
#MAGMOM = 92*2.5
#IVDW = 11
LCHARG = .F.
LWAVE = .F.
NPAR = 4
#KPAR = 4
KPOINTS：
auto mesh
0
G
4 4 4
0 0 0
作者Author: 嘿咻嘿咻跑run 时间: 2023-7-17 16:58
输入输出文件好像没有传上去
作者Author: Eudaimonia 时间: 2023-7-18 14:51
本帖最后由 Eudaimonia 于 2023-7-18 14:52 编辑
1.你查看一下cp2k的输出文件就知道cp2k在第一轮SCF收敛的时候收敛速度过慢了，跑了128次SCF才接近收敛限导致计算时长很高（而vasp首次SCF迭代才16步)；对inp文件的Mixing部分进行合理设置可以大幅度加快首轮SCF收敛(Alpha = 0.2, NBROYDEN = 12)，但是看第二轮SCF收敛需要20次反而不如Multiwfn的默认设置
2.合理
另外有个问题，我的配置的双路E5 2696V3,128G内存，提交36C运行，cp2k版本和你一致，为什么计算速度差50%~100%，我检查了CPU温度和内存，都有大量余裕
Ini.out 是直接用你的inp提交的，change.out是改动以后提交的
作者Author: 845417864 时间: 2023-7-18 20:50
cp2k印象里在大体系里用gamma点会比vasp快很多
作者Author: 乐平 时间: 2023-7-19 10:06
大体系的话， VASP 也可以用 vasp_gam 只考虑 1 1 1 Gamma 点
| 欢迎光临 计算化学公社 (http://bbs.keinsci.com/) |
Powered by Discuz! X3.3 |