|
采用vasp 6.4.2进行MLMD计算,温度设置为400K时进行计算比较顺利,在一定步数的从头计算之后会采用机器学习力场进行计算。
但是将温度设置为300K时,则会一直进行从头计算,直到1500步达到ML_MB的上限报错,但是在400K时,不到100步就开始使用机器学习力场进行预测。
各位大佬,小弟想请教一下为什么会出现更改温度后不进行机器学习力场预测这种问题?
以下是INCAR文件,300K和400K时仅对温度进行了修改。
- ALGO = Fast # Fast
- EDIFF = 0.00001
- ENCUT = 400
- ISMEAR = 1 #0
- SIGMA = 0.1 #0.05
- ISPIN = 2
- MAGMOM = 80*5.0 12*1.0
- ISTART = 1
- LREAL = Auto
- LWAVE = False
- NELM = 300
- NWRITE=1
- PREC = Normal
- NPAR = 4
- #FOR MD
- NSW=60000
- IBRION=0 #FOR MD
- POTIM=0.5
- ISIF = 2
- MDALGO = 4
- NHC_NCHAINS = 3
- NBLOCK=1
- KBLOCK=10
- NELMIN=6
- LCHARG =.F.
- TEBEG=300 # TEmperature at BEGining
- TEEND=300 #TEmperature at the END
- SMASS=1
- RANDOM_SEED = 12345678 0 0
- #Machine learning paramters
- ML_LMLFF = .TRUE.
- ML_ISTART = 0
- ML_LBASIS_DISCARD=.TRUE
复制代码 检查了一下300K时的ML_ABN文件,我已经跑了1000多步的从头计算,然而ML_ABN文件中却只有6个configurations,不知道这是什么原因。
此外对比了一下300K和400K时的ML_LOGFILE文件,在300K的计算在第二步时求解REGR用了50步,第8列的误差反而越来越大,不知道这个地方的不收敛该怎么解决。
在400K时,第二步的误差是逐渐下降的
- STATUS 0 threshold 2 T F 0 0
- --------------------------------------------------------------------------------
- STATUS 1 critical 4 T F 0 1
- LCONF 1 Fe 0 80 H 0 12
- SPRSC 1 1 1 Fe 80 25 H 12 12
- REGR 1 1 1 2.03427111E-01 5.58398988E-01 1.00179880E-09 7.32063669E+01
- REGR 1 1 2 6.54871649E-02 5.73960373E-01 3.19867735E-09 7.55316247E+01
- REGR 1 1 3 1.39153029E-02 5.89413409E-01 1.54586677E-08 7.78721935E+01
- REGR 1 1 4 1.40343056E-03 6.01123878E-01 1.56321159E-07 7.97374583E+01
- REGR 1 1 5 1.62097575E-04 6.06333461E-01 1.36514799E-06 8.07504988E+01
- REGR 1 1 6 6.10360115E-05 6.11595558E-01 3.65698258E-06 8.18132381E+01
- REGR 1 1 7 4.05461508E-05 6.17965968E-01 5.56236704E-06 8.28869727E+01
- REGR 1 1 8 3.35324992E-05 6.22468282E-01 6.77479283E-06 8.35984979E+01
- REGR 1 1 9 3.06198917E-05 6.25024219E-01 7.44968503E-06 8.39952530E+01
- REGR 1 1 10 2.93085681E-05 6.26357698E-01 7.79960364E-06 8.42008917E+01
- REGRF 1 1 11 2.86959729E-05 6.27026417E-01 7.97461269E-06 8.43037257E+01 1.66074114E+15 2.74002979E+09
- NDESC 1 24 Fe 452 H 452
- NDESC_SIC 1 Fe 226 H 226
- STDAB 1 0.00000000E+00 4.49911636E-02 0.00000000E+00
- ERR 1 1.33160414E-01 1.78565807E-02 1.50997415E+01
- CFE 1 0.00000000E+00 0.00000000E+00 0.00000000E+00
- LASTE 1 1.33160414E-01 1.78565807E-02 1.50997415E+01
- BEE 1 1.00000000E-06 3.46410162E-02 2.00000000E-02 2.00000000E-03 0.00000000E+00 0.00000000E+00
- BEEF 1 9.81859806E-04 1.20314673E-02 3.87184317E-03 2.00000000E-03 1.01482644E+00 6.42993152E-01
- --------------------------------------------------------------------------------
- STATUS 2 critical 4 T T 0 1
- LCONF 2 Fe 25 89 H 12 24
- SPRSC 2 2 2 Fe 89 31 H 24 24
- REGR 2 1 1 4.09510137E+00 6.83152741E-01 2.00000000E-15 1.83261423E+02
- REGR 2 1 2 1.78824428E+00 2.98318373E-01 2.00000000E-15 7.94947179E+01
- REGRF 2 1 3 1.79236326E+00 2.99005511E-01 2.00000000E-15 7.96778236E+01 -1.36179849E+16 8.34109683E+13
- NDESC 2 24 Fe 452 H 452
- NDESC_SIC 2 Fe 226 H 226
- STDAB 2 1.96774589E-05 6.13894063E-02 4.83205647E-02
- ERR 2 3.33592945E-06 2.83593686E-02 5.47773247E-02
- CFE 2 6.23424347E-02 2.51345630E-02 4.65148645E-01
- LASTE 2 3.33290111E-06 2.83506991E-02 5.78562090E-02
- BEE 2 9.81858332E-04 1.19683389E-02 3.87305125E-03 2.00000000E-03 1.01470925E+00 6.43202425E-01
- BEEF 2 1.00000000E-06 3.08795700E-02 6.41811580E-03 2.00000000E-03 2.73977324E-02 1.84484828E-02
复制代码
|
|