求助DLPNO-CCSD(T)内存问题

498746012

各位老师好，我的DLPNO-CCSD(T)开壳层计算总是报错，体系是50个原子加一个金属，服务器300G内存，分配了200G，内存已经挺大的了为什么还是报错呀？输入和输出文件如下：

! UKS  B3LYP/G def2-tzvp def2-tzvp/C def2/JK RIJK  DLPNO-CCSD(T) TightPNO SmallPrint Grid4 TightSCF nopop
%scf
    maxiter= 1000
    end

%maxcore 40000

%pal
    nprocs 5
    end

*xyz 0 3
输出文件
Generation of (ij|ab)[P] integrals            ... on
Generation of (ia|bc)[P],(ja|bc)[P] integrals ... on
Storage of 3 and 4 external integrals         ... on
Generation of ALL (ka|bc)[P] integrals        ... on
Keep RI integrals in memory                   ... off
Process 1:
Error (ORCA/RI_LocalTransformation): not enough memory to treat a single IJ-pair
Process 1: MemPerIJPair  1855.1 MB > MemAvailable -97383.6 MB

[file orca_mdci/mdci_dlpno_rhf_trafo.cpp, line 2949, Process 1]:   ... Aborting the run

[file orca_mdci/mdci_dlpno_rhf_trafo.cpp, line 2949, Process 1]:   ... Aborting the run

[localhost:309878] *** Process received signal ***
[localhost:309878] Signal: Aborted (6)
[localhost:309878] Signal code:  (-6)
[localhost:309878] [ 0] /lib64/libc.so.6[0x3d74a32510]
[localhost:309878] [ 1] /lib64/libc.so.6(gsignal+0x35)[0x3d74a32495]
[localhost:309878] [ 2] /lib64/libc.so.6(abort+0x175)[0x3d74a33c75]
[localhost:309878] [ 3] /home/orca421/liborca_tools_4_2_1_mpi.so(_Z16AbortMessageFunciPKciS0_i+0x92)[0x2b71fe2e9bf2]
[localhost:309878] [ 4] /home/orca421/orca_mdci_mpi[0x84e0c9]
[localhost:309878] [ 5] /home/orca421/orca_mdci_mpi[0x586249]
[localhost:309878] [ 6] /home/orca421/orca_mdci_mpi(main+0x501b)[0x14f5e1b]
[localhost:309878] [ 7] /lib64/libc.so.6(__libc_start_main+0xfd)[0x3d74a1ed1d]
[localhost:309878] [ 8] /home/orca421/orca_mdci_mpi[0x4d968d]
[localhost:309878] *** End of error message ***
--------------------------------------------------------------------------
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 1 with PID 0 on node localhost exited on signal 6 (Aborted).

498746012

wzkchem5 发表于 2021-3-17 17:48
用DFT轨道做DLPNO一般是因为HF reference定性错误但是KS reference定性正确，此时用KS轨道做参考态结果更 ...

谢谢老师的耐心讲解

wzkchem5

niobium 发表于 2021-3-17 11:32
可能只能慢慢减少核数，加大内存了。最近看到不少基于DFT做DLPNO的，这样有依据吗？结果更好？
你是做什么 ...

用DFT轨道做DLPNO一般是因为HF reference定性错误但是KS reference定性正确，此时用KS轨道做参考态结果更靠谱。KS轨道和HF轨道的差别基本上会被CCSD(T)的singles cluster operator修正掉

498746012

niobium 发表于 2021-3-17 11:32
可能只能慢慢减少核数，加大内存了。最近看到不少基于DFT做DLPNO的，这样有依据吗？结果更好？
你是做什么 ...

谢谢老师，我刚接触orca，没有任何经验，没有什么依据

niobium

可能只能慢慢减少核数，加大内存了。最近看到不少基于DFT做DLPNO的，这样有依据吗？结果更好？
你是做什么问题的，如果不是算弱相互作用，用NormalPNO的话，精度也说得过去的。内存应该会小很多。