计算化学公社

标题: orca 5.0 不能用RI加速？ [打印本页]

作者
Author: huozhong 时间: 2021-7-13 18:03
标题: orca 5.0 不能用RI加速？
有没有遇到类似情况的？是ORCA5.0不能用RI加速吗？【至少以下这种方式不行，还是哪里搞错了】

刚安装orca5.0 测试H2分子的优化和频率分析，没有问题。----------------------------------------------------------------
[root@localhost oi]# mpiexec --versionmpiexec (OpenRTE) 4.1.1

Report bugs to http://www.open-mpi.org/community/help/

[root@localhost oi]# which orca
alias orca='/root/hcp/orca500/orca'
/root/hcp/orca500/orca

输入文件用multiwfn 转化的inp格式：
! B97-3c opt freq tightSCF noautostart miniprint nopop
%maxcore  5000
%pal nprocs 7 end
* xyz 0 1
H    -3.24303100 0.26387009 0.00000000
H    -3.84303100 0.26387009 0.00000000
*

输出结果：
Timings for individual modules:

Sum of individual times       ...       4.248 sec (= 0.071 min)
GTO integral calculation       ...       0.857 sec (= 0.014 min)  20.2 %
SCF iterations                ...       1.111 sec (= 0.019 min)  26.1 %
SCF Gradient evaluation       ...       1.218 sec (= 0.020 min)  28.7 %
Geometry relaxation          ...       0.055 sec (= 0.001 min) 1.3 %
Analytical frequency calculation...       1.007 sec (= 0.017 min)  23.7 %
                           ****ORCA TERMINATED NORMALLY****
TOTAL RUN TIME: 0 days 0 hours 0 minutes 5 seconds 106 msec

-----------------------------------------------------------------------------------

但是在用multiwfn中的方法： 5 RI-wB97M-V/def2-TZVP或者
6 RI-PWPB95-D3(BJ)/def2-TZVPP时，计算就会报错，【其他设置都不变】请问这是怎么回事？
----------------------------------------------------------------------------------
报错的输入文件：
! wB97M-V def2-TZVP def2/J RIJCOSX strongSCF opt freq tightSCF noautostart miniprint nopop
%maxcore  5000
%pal nprocs 7 end
* xyz 0 1
H    -3.24303100 0.26387009 0.00000000
H    -3.84303100 0.26387009 0.00000000
*

报错的输出文件：
WARNING: Gradients are needed
  ===> : At least TightSCF should be used for SCF convergence

WARNING: Analytic gradients for DFT-NL dispersion can only be done using the self-consistent method,
  ===> Switching to SCNL

WARNING: DFT-NL dispersion correction analytic Hessian not yet implemented:
      Please use !NUMFREQ instead!
  ===> Skipping actual calculation

[file orca_main/mainchk.cpp, line 4930]: Error (ORCA_MAIN): ... aborting the run

----------------------------------------------------------------------------------
本来是打算用RI-wB97M-V/def2-TZVP优化有机镍配合物的结构和频率分析（101个原子，含Ni, N, Si, C, H元素），发现报错，就用该方法尝试了氢气分子的优化和频率分析，发现也报错，但是用B97-3c时就可以正常结束。
----------------------------------------------------------------------------------
优化有机镍配合物时的输入信息：
! CCSD(T) cc-pVTZ tightSCF opt freq noautostart miniprint nopop
%maxcore  5000
%pal nprocs 7 end
* xyz 0 1
Ni -2.53594800 -0.32917300 0.25771000

......
H    2.60528200 -0.51459700 -3.27719000
*

优化有机镍配合物时的报错信息：
Reading SHARK input file Nialkyl.SHARKINP.tmp ... ok
----------------------
SHARK INTEGRAL PACKAGE
----------------------

Number of atoms                            ... 101
Number of basis functions                ... 1751
Number of shells                         ... 725
Maximum angular momentum                   ...    3
Integral batch strategy                   ... SHARK/LIBINT Hybrid
RI-J (if used) integral strategy          ... SPLIT-RIJ (Revised 2003 algorithm where possible)
Printlevel                               ...    1
Contraction scheme used                   ... SEGMENTED contraction
Coulomb Range Separation                   ... NOT USED
Exchange Range Separation                ... NOT USED
Finite Nucleus Model                      ... NOT USED
Auxiliary Coulomb fitting basis          ... AVAILABLE
# of basis functions in Aux-J          ... 2863
# of shells in Aux-J                   ... 965
Maximum angular momentum in Aux-J       ...    4
Auxiliary J/K fitting basis                ... NOT available
Auxiliary Correlation fitting basis       ... NOT available
Auxiliary 'external' fitting basis       ... NOT available
Integral threshold                         ...    2.500000e-11
Primitive cut-off                         ...    2.500000e-12
Primitive pair pre-selection threshold    ...    2.500000e-12

Calculating pre-screening integrals       ... done (  0.9 sec) Dimension = 725
Organizing shell pair data                ... done (  0.8 sec)
Shell pair information
Total number of shell pairs                ... 263175
Shell pairs after pre-screening          ... 158668
Total number of primitive shell pairs    ... 794379
Primitive shell pairs kept                ... 297151
      la=0 lb=0:  46522 shell pairs
      la=1 lb=0:  47961 shell pairs
      la=1 lb=1:  12467 shell pairs
      la=2 lb=0:  21248 shell pairs
      la=2 lb=1:  11081 shell pairs
      la=2 lb=2: 2514 shell pairs
      la=3 lb=0: 9327 shell pairs
      la=3 lb=1: 4925 shell pairs
      la=3 lb=2: 2149 shell pairs
      la=3 lb=3: 474 shell pairs

Calculating one electron integrals       ... done (  0.4 sec)
Calculating RI/J V-Matrix + Cholesky decomp....
ORCA finished by error termination in GTOInt
Calling Command: mpirun -np 7  /root/hcp/orca500/orca_gtoint_mpi Nialkyl.int.tmp Nialkyl
[file orca_tools/qcmsg.cpp, line 458]:
  .... aborting the run
----------------------------------------------------------------------------------

作者
Author: hebrewsnabla 时间: 2021-7-13 18:08
本帖最后由 hebrewsnabla 于 2021-7-13 18:13 编辑

wB97M-V不支持Hessian，当然会报错了

作者
Author: huozhong 时间: 2021-7-13 18:09
【更正】
‘优化有机镍配合物时的输入信息：
! CCSD(T) cc-pVTZ tightSCF opt freq noautostart miniprint nopop’

应该是：

‘优化有机镍配合物时的输入信息：
! wB97M-V def2-TZVP def2/J RIJCOSX strongSCF opt freq tightSCF noautostart miniprint nopop’

作者
Author: huozhong 时间: 2021-7-13 18:12

hebrewsnabla 发表于 2021-7-13 18:08
Multiwfn生成的输入文件没有opt freq吧，这是你自己加的。

wB97M-V不支持Hessian，当然会报错了

刚才输入文件写错了，抱歉，已在楼下更正。
Multiwfn可以生成输入文件opt freq
4 Optimization + Frequency

作者
Author: wzkchem5 时间: 2021-7-13 18:12
WARNING: DFT-NL dispersion correction analytic Hessian not yet implemented:
Please use !NUMFREQ instead!
===> Skipping actual calculation

输出文件写这么大字你看不见吗，这根本和RI没关系

作者
Author: 喵星大佬 时间: 2021-7-13 18:17

wzkchem5 发表于 2021-7-13 18:12
WARNING: DFT-NL dispersion correction analytic Hessian not yet implemented:
Please use !NU ...

not yet implemented

总给人一种很快会实现的错觉

作者
Author: akber123 时间: 2021-7-14 14:14
检查一下，你机子有没有安装gfortran，如果是没安装的话，安装下，然后重新编译openmpi试试

作者
Author: zjxitcc 时间: 2021-8-22 23:57

akber123 发表于 2021-7-14 14:14
检查一下，你机子有没有安装gfortran，如果是没安装的话，安装下，然后重新编译openmpi试试

他这不是编译器和库的问题，是用了一个不支持（还没实现）的功能。

欢迎光临计算化学公社 (http://bbs.keinsci.com/)