[VASP]电子迭代的收敛问题
提交者: avan 日期: 2006/11/5 13:16 阅读: 216
作者:avan
来源:量子化学网
摘要:对于一些磁性体系、镧系和锕系元素及相关化合物的静态计算（电子迭代），经常会遇到“难收敛”的问题。
下面给出几个相关Flag及设置方法：（有些话从说明书上摘录的，不大好翻译，就copy了，请见谅）
Tag: VASP 软件经验 电子迭代 疑难解答 [作者注] 目前这文章看起来有点乱，过段时间我会再重新整理。

  1、LMAXMIX
   Default: LMAXMIX = 2
   An additional flag controls up to which l quantum number the onsite PAW charge densities are passed through the charge density mixer. Higher l-quantum numbers are usually not handled by the mixer. In order to obtain fast convergence to the groundstate, you can try the following setting:
  
   LMAXMIX = 4 for d elements
   LMAXMIX = 6 for f elements
  
   这个FLAG对于含d电子和f电子的体系是非常重要的，很大一部分体系的收敛问题可以通过设置合适的LMAXMIX值来解决。

  2、ALGO, IALGO, LDIAG
   If the self-consistency loop does not converge within 40 steps, it will probably not converge at all. In this case you should reconsider the tags IALGO, LDIAG, and the mixing-parameters.
   这是说明书上的建议。
   一般情况下，或使用IALGO=48时遇到收敛问题的话，可以考虑设IALGO为38（4.5以前的版本可设为8），或设置ALGO=Normal or Fast (in VASP.4.5 and later versions)。

  3、NELMDL
   NELMDL gives the number of non-selfconsistent steps at the beginning; if one initializes the wave functions randomly the initial wave functions are far from anything reasonable. The resulting charge density is also 'nonsense'. Therefore it makes sense to keep the initial Hamiltonian, which corresponds to the superposition of atomic charge densities, fixed during the first few steps.
  Choosing a 'delay' for starting the charge density update becomes essential in all cases where the SC-convergence is very bad (e.g. surfaces or molecules/clusters chains). Without setting a delay VASP will probably not converge or at least the convergence speed is slowed down.
  NELMDL might be positive or negative. A positive number means that a delay is applied after each ionic movement -- in general not a convenient option. A negative value results in a delay only for the start-configuration.

  4、mixing-parameters
   对于一些难收敛的体系，可以使用“linear mixing”,具体详见VASP说明书中的“Mixing-tags”。

   For an initial linear mixing (BMIX ~ 0) an optimal setting for A(AMIX) can be found easily by setting Aopt=Acurrent*Γmean. For the Kerker scheme either A or q0(i.e. AMIX or BMIX) can be optimized, but we recommend to change only BMIX and keep AMIX fixed (you must decrease BMIX if the mean eigenvalue is larger than one, and increase BMIX if the mean eigenvalue is smaller than one).
   
   尽管VASP说明书中给出了调节AMIX和BMIX的一些较为明确的建议，但是实际去调节的时候，还是挺难的，但原则上说，是可以通过调节这两个Flag来使得收敛问题得以解决的，只是得有耐心。

  5、kmesh, SIGMA
   收敛问题还跟kmesh及SIGMA(当使用ISMEAR不等于-5 和-4时)的设置有关。要达到同样的精度，较小的SIGMA则需要较大的kmesh；而且，当SIGMA较小时，若kpoints不够多，也会出现难收敛的情况。

   P.S. 对于不同系统的计算，问题的原因不一定一样，因而可能解决之一问题的方法也不一定会一样。其实以上大部分都来自VASP的说明书，只是我在遇到收敛问题时试过这些方法.

   写得比较仓促，以后再修改及补充。
   希望有这方面经验的朋友也把自己的见解提出来一起讨论，

看看英文版的手册P76页,也许有帮助
AMIX,BMIX等