计算化学公社

标题: ORCA计算164个原子的有机大分子频率出错 [打印本页]

作者
Author: piggyFLYING 时间: 2023-8-5 09:57
标题: ORCA计算164个原子的有机大分子频率出错
我使用ORCA5.0.0计算164个原子的有机大分子（如下图）几何优化、频率，

输入文件：
! B3LYP def2-SVP def2/J RIJCOSX opt freq noautostart miniprint nopop
%maxcore 16000
%pal nprocs 12 end
* xyz 0 1

...................
在计算过程中出现以下问题：

-------------------------------------------------------------------------------
                           ORCA SCF HESSIAN
-------------------------------------------------------------------------------

Hessian of the Kohn-Sham DFT energy:
Kohn-Sham wavefunction type                   ... RKS
Hartree-Fock exchange scaling                   ... 0.200
Number of operators                            ... 1
Number of atoms                               ...  164
Basis set dimensions                            ... 1956
Integral neglect threshold                      ... 2.5e-11
Integral primitive cutoff                      ... 2.5e-12

Setting up DFT Hessian calculations             ...

HESS GRID
---------

General Integration Accuracy    IntAcc    ... 4.629
Radial Grid Type                RadialGrid  ... OptM3 with GC (2021)
Angular Grid (max. ang.)       AngularGrid ... 5 (Lebedev-434)
Angular grid pruning method    GridPruning ... 4 (adaptive)
Weight generation scheme       WeightScheme... Becke
Basis function cutoff          BFCut    ... 1.0000e-11
Integration weight cutoff       WCut       ... 1.0000e-14
Angular grids for H and He will be reduced by one unit
Partially contracted basis set             ... off
Rotationally invariant grid construction    ... off

Total number of grid points                ...  1341686
Total number of batches                   ... 21038
Average number of points per batch          ...    63
Average number of grid points per atom    ...    8181

Building densities                            ... done (    1.2 sec)
Calculating rho(r) on the grid                ... done (    4.2 sec)
Building xc-kernel on the grid                ... done (    0.3 sec)
                                                   done (    31.9 sec)

Nuclear repulsion Hessian  (SHARK)             ... done (    0.0 sec)

----------------------------------------------
Forming right-hand sides of CP-SCF equations    ...
----------------------------------------------
One electron integral derivatives  (SHARK)    ... done (    17.5 sec)
Transforming the overlap derivative matrices    ... done (    26.6 sec)
Making the Q(x) pseudodensities                ... done (    26.5 sec)
Adding the E*S(x)*S(y) terms to the Hessian    ... done ( 152.8 sec)
Calculating energy weighted overlap derivatives  ... done (    12.1 sec)
Two electron integral derivatives (SHARK-RI)    ... done ( 1347.0 sec)
Exchange-correlation integral derivatives       ... done ( 285.6 sec)
tr(F(y)Q(x)) contribution to the Hessian       ... done ( 209.0 sec)

Response fock operator R(S(x)) (SHARK-RIJCOSX) ...
IBatch  1 (of  2):                                  --------------------------------------------------------------------------
Primary job  terminated normally, but 1 process returned
a non-zero exit code. Per user-direction, the job has been aborted.
--------------------------------------------------------------------------
--------------------------------------------------------------------------
mpirun noticed that process rank 0 with PID 80078 on node node116 exited on signal 9 (Killed).
--------------------------------------------------------------------------

ORCA finished by error termination in SCF Hessian
Calling Command: mpirun -np 12  /share/apps/orca/orca_5_0_0_linux_x86-64_shared_openmpi411/orca_scfhess_mpi P_single.scfhess.inp P_single
[file orca_tools/qcmsg.cpp, line 458]:
  .... aborting the run

[file orca_tools/qcmsg.cpp, line 458]:
  .... aborting the run

请问各位老师如何解决？

作者
Author: wzkchem5 时间: 2023-8-5 15:36
大概率是内存不足

作者
Author: dnlx 时间: 2023-8-7 03:50
柔性大分子做振动分析，B3LYP显然应该加D3。

另外这个分子有太多的共轭pi键，B3LYP可能不适合做优化+振动分析
("B3LYP不适合优化大共轭体系，这点很多人都不知道。"http://sobereva.com/557)

但是wB97X-D3更费时间和内存

作者
Author: KSeGaSn 时间: 2023-8-7 09:37

wzkchem5 发表于 2023-8-5 15:36
大概率是内存不足

老师您好，我在做一个和楼主体系大小相似的频率计算时，到了Response fock operator R(S(x)) (SHARK-RIJCOSX)分23个Batch，顺利计算了15个之后在第16个batch上卡死，五倍于上一个Batch计算时间里out文件无更新，CPU和内存都不再被大量占用，请问这个是什么问题呢？是不是也只能重新提交任务计算了，谢谢老师。

作者
Author: wzkchem5 时间: 2023-8-7 16:02

KSeGaSn 发表于 2023-8-7 02:37
老师您好，我在做一个和楼主体系大小相似的频率计算时，到了Response fock operator R(S(x)) (SHARK-RIJC ...

是不是硬盘满了？

作者
Author: KSeGaSn 时间: 2023-8-7 21:49

wzkchem5 发表于 2023-8-7 16:02
是不是硬盘满了？

没有，我特意留了900G的硬盘算这个频率，但是只写入了60G的数据。我一直放到现在也没有任何报错，我去论坛上也没看到什么类似的情况，可能是bug？另外老师我是做一个优化后T1态的虚频检验（UKS），光算这个fock operator就要用9000s * 23Batch，是不是到后面CP-SCF也不太能算得动啊……

作者
Author: wzkchem5 时间: 2023-8-7 22:44

KSeGaSn 发表于 2023-8-7 14:49
没有，我特意留了900G的硬盘算这个频率，但是只写入了60G的数据。我一直放到现在也没有任何报错，我去论 ...

看看是不是maxcore设置不合理，maxcore太小会导致batch数增加、计算耗时增加。至于计算不再更新的问题，也可能是openmpi设置有问题，或者遇到了偶然的硬件问题

欢迎光临计算化学公社 (http://bbs.keinsci.com/)