|
本帖最后由 djjj148 于 2021-2-21 16:58 编辑
最近在学习CP2K,尝试了一个89原子的体系作为练习,发现CP2K跑一步CI-NEB过渡态的离子步需要多次scf收敛,最重要的是这个这个“多次”的次数还不固定,少则3-4次,多则十几次。同样的体系做普通的结构优化3-4小时就能跑150多步离子步并收敛,然而过渡态计算在花了4小时内只跑了两步离子步!我插了2个点,一些具有代表性的信息筛选如下:
简单来说,现在CI-NEB计算就跑了2步离子步,scf收敛了4次才跑的第一步,然后scf收敛了3次才进入第二步离子步,但是之后scf收敛了12次都没有进入第三步离子步。- $ cat -n cp2k-BAND4.out |tail -1
- 6134
复制代码- $ grep "RMS DISPLACEMENT" cp2k-BAND4.out -n -A4
- 1078: RMS DISPLACEMENT = 0.01891 [ 0.02000] (YES)
- 1079- MAX DISPLACEMENT = 0.28522 [ 0.01000] ( NO)
- 1080- RMS FORCE = 0.00664 [ 0.00100] ( NO)
- 1081- MAX FORCE = 0.06832 [ 0.00100] ( NO)
- 1082- *******************************************************************************
- --
- 1979: RMS DISPLACEMENT = 0.01105 [ 0.02000] (YES)
- 1980- MAX DISPLACEMENT = 0.11364 [ 0.01000] ( NO)
- 1981- RMS FORCE = 0.00470 [ 0.00100] ( NO)
- 1982- MAX FORCE = 0.05926 [ 0.00100] ( NO)
- 1983- *******************************************************************************
复制代码- $ sed -n "1,1082p" cp2k-BAND4.out | grep "SCF run converged" | cat -n
- 1 *** SCF run converged in 82 steps ***
- 2 *** SCF run converged in 67 steps ***
- 3 *** SCF run converged in 37 steps ***
- 4 *** SCF run converged in 94 steps ***
复制代码- $ sed -n "1083,1983p" cp2k-BAND4.out | grep "SCF run converged" | cat -n
- 1 *** SCF run converged in 13 steps ***
- 2 *** SCF run converged in 16 steps ***
- 3 *** SCF run converged in 7 steps ***
复制代码- $ sed -n "1984,6134p" cp2k-BAND4.out | grep "SCF run converged" | cat -n
- 1 *** SCF run converged in 25 steps ***
- 2 *** SCF run converged in 88 steps ***
- 3 *** SCF run converged in 88 steps ***
- 4 *** SCF run converged in 22 steps ***
- 5 *** SCF run converged in 85 steps ***
- 6 *** SCF run converged in 70 steps ***
- 7 *** SCF run converged in 10 steps ***
- 8 *** SCF run converged in 85 steps ***
- 9 *** SCF run converged in 25 steps ***
- 10 *** SCF run converged in 49 steps ***
- 11 *** SCF run converged in 67 steps ***
- 12 *** SCF run converged in 61 steps ***
复制代码
而且我翻看输出的cp2k-BAND4.out和其他文件也没有发现异常,这样的过渡态计算效率实在太低了,想请教大家:
为什么CP2K跑一步CI-NEB过渡态的离子步需要多次scf收敛,而且这个收敛次数还不固定呢?
inp文件和4个点的out文件如下,还请大家指教一下,谢谢!
|
|