怎样才能模拟较大样本的原子位置

ZRItchi

组里想模拟样本在不同状态下(液体、声子) 被x光照射后的散射speckle pattern， 然后看能否从这些随时间变化的图像中逆推回原本的样本状态。首先想要模拟的是锗在室温(晶体) 和液态情况下的原子位置，样本大小大约是100 x 100 x 50 nm。使用LAMMPS进行小规模计算后，发现有声子谱很多虚频，且结果很大程度受选用的势影响。
于是我们打算使用Quantum Espresso 的 ab initio md 来进行更精确的模拟。可是在能够查找到的资料和范例里，都没有发现如何提取每个step 的原子位置。而且使用md 方法对 2 x 2 x 2 个 unit cell (64个原子) 进行计算时，estimated dynamical ram 就已经有200多G了。这是因为QE md 的计算过程中，"cutoff distance" 很长吗?

请问各位老师，对于这种计算，有没有什么推荐的工具。如果不通过模拟，有没有什么好的方法来近似这些原子的位置呢？


------------- QE input file -----------------
&control
    calculation='md'
    restart_mode='from_scratch',
    prefix = 'Ge',
    pseudo_dir = './pseudo/',
    !outdir='./output/',
    dt=20,
    nstep=10,
    disk_io='low'
/
&system
    ibrav= 0,
    nat=  64, ntyp= 1,
    ecutwfc = 60.0,  ! this needs to be tuned for a convergence study
    ecutrho = 600.0,
    nosym=.true.
/
&electrons
    conv_thr =  1.0d-6
    mixing_beta = 0.7
/
&ions
    pot_extrapolation='second-order'
    wfc_extrapolation='second-order'
/
ATOMIC_SPECIES
Ge  72.64  Ge.pbe-dn-kjpaw.UPF
K_POINTS crystal_b
6
0.0   0.0    0.0   100
0.0   0.5    0.0   100
0.0   0.375 -0.375 100
-0.25 0.25  -0.5   100
-0.5  0.0   -0.5   100
0.0   0.0    0.0   1
!   ibrav = 0, nat = 64, ntyp = 1
CELL_PARAMETERS angstrom
   11.3159999999999989 0.0000000000 0.0000000000
   0.0000000000 11.3159999999999989 0.0000000000
   0.0000000000 0.0000000000 11.3159999999999989
ATOMIC_POSITIONS alat
Ge  0.0000000000000000  0.0000000000000000  0.0000000000000000
Ge  0.0000000000000000  0.2500000000000000  0.2500000000000000
Ge  0.2500000000000000  0.0000000000000000  0.2500000000000000
Ge  0.2500000000000000  0.2500000000000000  0.0000000000000000
Ge  0.3750000000000000  0.3750000000000000  0.3750000000000000
Ge  0.3750000000000000  0.1250000000000000  0.1250000000000000
Ge  0.1250000000000000  0.3750000000000000  0.1250000000000000
Ge  0.1250000000000000  0.1250000000000000  0.3750000000000000
Ge  0.0000000000000000  0.0000000000000000  0.5000000000000000
Ge  0.0000000000000000  0.2500000000000000  0.7500000000000000
Ge  0.2500000000000000  0.0000000000000000  0.7500000000000000
Ge  0.2500000000000000  0.2500000000000000  0.5000000000000000
Ge  0.3750000000000000  0.3750000000000000  0.8750000000000000
Ge  0.3750000000000000  0.1250000000000000  0.6250000000000000
Ge  0.1250000000000000  0.3750000000000000  0.6250000000000000
Ge  0.1250000000000000  0.1250000000000000  0.8750000000000000
Ge  0.0000000000000000  0.5000000000000000  0.0000000000000000
Ge  0.0000000000000000  0.7500000000000000  0.2500000000000000
Ge  0.2500000000000000  0.5000000000000000  0.2500000000000000
Ge  0.2500000000000000  0.7500000000000000  0.0000000000000000
Ge  0.3750000000000000  0.8750000000000000  0.3750000000000000
Ge  0.3750000000000000  0.6250000000000000  0.1250000000000000
Ge  0.1250000000000000  0.8750000000000000  0.1250000000000000
Ge  0.1250000000000000  0.6250000000000000  0.3750000000000000
Ge  0.0000000000000000  0.5000000000000000  0.5000000000000000
Ge  0.0000000000000000  0.7500000000000000  0.7500000000000000
Ge  0.2500000000000000  0.5000000000000000  0.7500000000000000
Ge  0.2500000000000000  0.7500000000000000  0.5000000000000000
Ge  0.3750000000000000  0.8750000000000000  0.8750000000000000
Ge  0.3750000000000000  0.6250000000000000  0.6250000000000000
Ge  0.1250000000000000  0.8750000000000000  0.6250000000000000
Ge  0.1250000000000000  0.6250000000000000  0.8750000000000000
Ge  0.5000000000000000  0.0000000000000000  0.0000000000000000
Ge  0.5000000000000000  0.2500000000000000  0.2500000000000000
Ge  0.7500000000000000  0.0000000000000000  0.2500000000000000
Ge  0.7500000000000000  0.2500000000000000  0.0000000000000000
Ge  0.8750000000000000  0.3750000000000000  0.3750000000000000
Ge  0.8750000000000000  0.1250000000000000  0.1250000000000000
Ge  0.6250000000000000  0.3750000000000000  0.1250000000000000
Ge  0.6250000000000000  0.1250000000000000  0.3750000000000000
Ge  0.5000000000000000  0.0000000000000000  0.5000000000000000
Ge  0.5000000000000000  0.2500000000000000  0.7500000000000000
Ge  0.7500000000000000  0.0000000000000000  0.7500000000000000
Ge  0.7500000000000000  0.2500000000000000  0.5000000000000000
Ge  0.8750000000000000  0.3750000000000000  0.8750000000000000
Ge  0.8750000000000000  0.1250000000000000  0.6250000000000000
Ge  0.6250000000000000  0.3750000000000000  0.6250000000000000
Ge  0.6250000000000000  0.1250000000000000  0.8750000000000000
Ge  0.5000000000000000  0.5000000000000000  0.0000000000000000
Ge  0.5000000000000000  0.7500000000000000  0.2500000000000000
Ge  0.7500000000000000  0.5000000000000000  0.2500000000000000
Ge  0.7500000000000000  0.7500000000000000  0.0000000000000000
Ge  0.8750000000000000  0.8750000000000000  0.3750000000000000
Ge  0.8750000000000000  0.6250000000000000  0.1250000000000000
Ge  0.6250000000000000  0.8750000000000000  0.1250000000000000
Ge  0.6250000000000000  0.6250000000000000  0.3750000000000000
Ge  0.5000000000000000  0.5000000000000000  0.5000000000000000
Ge  0.5000000000000000  0.7500000000000000  0.7500000000000000
Ge  0.7500000000000000  0.5000000000000000  0.7500000000000000
Ge  0.7500000000000000  0.7500000000000000  0.5000000000000000
Ge  0.8750000000000000  0.8750000000000000  0.8750000000000000
Ge  0.8750000000000000  0.6250000000000000  0.6250000000000000
Ge  0.6250000000000000  0.8750000000000000  0.6250000000000000
Ge  0.6250000000000000  0.6250000000000000  0.8750000000000000

ZRItchi

补充一下。如果有办法从 QE scf + phonopy 的声子计算结果中，提取每个mode的振动方向和振幅，是否可以手动把每个原子的位置套在多个简谐振动里，然后添加某种noise来模拟原子的位置呢？

get-it

啥叫“原本的样本状态”

ZRItchi

get-it 发表于 2022-3-12 21:50
啥叫“原本的样本状态”

大概就是通过看 speckle pattern 的变化反推出样本到底是 liquid 还是 crystal 或者 glass 之类的。现在的diagnostics 只有看speckle contrast, 然而光看 contrast 得到的结论很局限。

ZRItchi

经过一段时间的讨论，我们放弃通过计算电子 (ab-initio md) 来获取原子位置了。接下来我们打算用 LAMMPS 粗略地模拟一下看看，如果有 semi-reasonable 的dispersion curve 的话就拿来用了。