求助：lammps运行时报错not enough space for bonds

灰天皮匠

各位前辈好，我在用lammps反应力场的模块跑动力学，系综是npt，体系是金属和H2气氛，但是报错not enough space for bonds。在Mailing List上查了一下，尝试了减少时间步长和减少核数运行这两种方法，发现问题依然解决不了（只有一个核时可以算，但是速度非常慢，目前还不知道会不会正常终止）。不知道是否有人碰上过同样的问题，以及该如何解决，恳请大佬们赐教

ddls2017

同问，我也遇到了这个问题？请问楼主这个问题解决了吗？怎么解决的？多谢！

liucw

报错信息如下：

p17: not enough space for bonds! total=74916 allocated=74050

application called MPI_Abort(MPI_COMM_WORLD, -14) - process 17

p23: not enough space for bonds! total=96532 allocated=95161

application called MPI_Abort(MPI_COMM_WORLD, -14) - process 23

在mailing list中查了一下，尝试用5个核可以运行。

Steve的回复是：

This is a question for Metin. Usually a memory problem indicates something bogus about the way you have set up your model.

并且Metin提出建议：

1) As Steve pointed out, there may be a problem with your time-step parameter. The fact that reax/c runs out of space for bonds shows that your system is evolving rather too quickly (and not doing meaningful physics). I would suggest that you decrease your time-step length.

2) If you are certain about the physics done with your current set of parameters, but think that reax/c is not able to keep up with your system, you may try increasing the MIN_BONDS, SAFE_ZONE and SAFER_ZONE parameters defined in reaxc_defs.h and re-run your simulation after re-compilation. However, I strongly recommend the first option, not this one.