求助：用cp2k使用SCCS情况下做结构优化难收敛的问题

Wenlin

求助大家，我正在计算一个COF吸附锕酰离子的结构，因为要考虑实际的溶液环境所以必须考虑使用SCCS，尝试了调节很多关于SCF以及SCCS的参数，有收敛趋势但是还是差很多，主要是想问下大家关于SCCS的参数设置有没有哪里可以调整的，而且我看训练班的ppt关于SCCS部分的描述有个EPS_SCF的参数，一般默认0.5，这个值和平时常设的SCF部分EPS_SCF设置的有什么不同呢？因为一般来讲SCF部分好像都是设置1E-6或者5E-6，这里的0.5是不是太大了？我不太理解SCCS部分EPS_SCF默认值为0.5的意义，下面是我的输入文件，麻烦老师看一下有没有需要调整的或者能够帮助收敛的。


&DFT
    BASIS_SET_FILE_NAME  BASIS_MOLOPT_AcPP1
           BASIS_SET_FILE_NAME  BASIS_MOLOPT
    POTENTIAL_FILE_NAME  POTENTIAL
#   WFN_RESTART_FILE_NAME SUPERCOFdisp-AmO2^+-RESTART.wfn
    CHARGE    1 #Net charge
    MULTIPLICITY    3 #Spin multiplicity
    UKS
    PLUS_U_METHOD Lowdin #The method used in DFT+U. Can also be Lowdin
    &QS
      EPS_DEFAULT 1.0E-12 #Set all EPS_xxx to values such that the energy will be correct up to this value
    &END QS
    &POISSON
      PERIODIC XY #Direction(s) of PBC for calculating electrostatics
      PSOLVER MT #The way to solve Poisson equation
    &END POISSON
    &XC
      &XC_FUNCTIONAL PBE
      &END XC_FUNCTIONAL
    &END XC
    &MGRID
      CUTOFF  500
      REL_CUTOFF  60
    &END MGRID
    &SCF
      MAX_SCF 25 #Maximum number of steps of inner SCF
      EPS_SCF 5.0E-06 #Convergence threshold of density matrix of inner SCF
#     SCF_GUESS RESTART #Use wavefunction from WFN_RESTART_FILE_NAME file as initial guess
#     IGNORE_CONVERGENCE_FAILURE #Continue calculation even if SCF not converged, works for version >= 2024.1
      &OT
        PRECONDITIONER FULL_SINGLE_INVERSE #FULL_SINGLE_INVERSE is also worth to try. FULL_ALL is better but quite expensive for large system
        MINIMIZER CG #CG is worth to consider in difficult cases
        LINESEARCH 2PNT #1D line search algorithm for CG. 2PNT is default. 3PNT is more expensive but may be better. GOLD is best but very expensive
        ALGORITHM IRAC #Algorithm of OT. Can be STRICT (default) or IRAC
      &END OT
      &OUTER_SCF
        MAX_SCF 20 #Maximum number of steps of outer SCF
        EPS_SCF 5.0E-06 #Convergence threshold of outer SCF
      &END OUTER_SCF
      &PRINT
        &RESTART #Note: Use "&RESTART OFF" can prevent generating .wfn file
          BACKUP_COPIES 0 #Maximum number of backup copies of wfn file. 0 means never
        &END RESTART
      &END PRINT
    &END SCF
    &PRINT
      #Uncomment following lines can print occupation for which +U is applied when PRINT_LEVEL is medium
      #&PLUS_U
      #  &EACH
      #    QS_SCF 1
      #  &END EACH
      #&END PLUS_U
      @IF 1 #Printing SCCS information in each SCF iteration
      &SCCS
        &EACH
          QS_SCF 1
        &END EACH
        &POLARISATION_CHARGE_DENSITY
          &EACH
            QS_SCF 0
          &END EACH
          STRIDE 2
        &END POLARISATION_CHARGE_DENSITY
        &DIELECTRIC_FUNCTION
          &EACH
            QS_SCF 0
          &END EACH
          STRIDE 2
        &END DIELECTRIC_FUNCTION
      &END SCCS
      @ENDIF
    &END PRINT
    &SCCS
      ALPHA [mN/m] 57.2
      BETA [GPa] -0.5
      GAMMA [mN/m] 0.0
      DIELECTRIC_CONSTANT 78.36
      EPS_SCF 0.5 #SCCS is activated only if SCF iteration is converged to this threshold
      EPS_SCCS 1E-8 #Requested accuracy for convergence of polarization charge density iteration
      MAX_ITER 200 #Maximum number of polarization charge density iterations
      DERIVATIVE_METHOD CD5 #Method for calculation of numerical derivatives. Can be FFT, CD3, CD5, CD7
      &ANDREUSSI
        RHO_MIN 0.0001
        RHO_MAX 0.001
      &END ANDREUSSI
    &END SCCS
  &END DFT
&END FORCE_EVAL

&MOTION
  &GEO_OPT
    TYPE MINIMIZATION #Search for minimum
    KEEP_SPACE_GROUP F #If T, then space group will be detected and preserved
    OPTIMIZER LBFGS #Can also be CG (more robust for difficult cases) or BFGS
    &LBFGS
      TRUST_RADIUS 0.05 #Trust radius (maximum stepsize) in Angstrom
      MAX_H_RANK 5 #Larger values (e.g. 30) will accelerate convergence behaviour at the cost of a larger memory consumption
    &END LBFGS
    MAX_ITER 500 #Maximum number of geometry optimization
    MAX_DR 3E-3 #Maximum geometry change
    RMS_DR 1.5E-3 #RMS geometry change
    MAX_FORCE 4.5E-4 #Maximum force
    RMS_FORCE 3E-4 #RMS force
  &END GEO_OPT
  &PRINT
    &TRAJECTORY
      FORMAT xyz
    &END TRAJECTORY
    &RESTART
      BACKUP_COPIES 0 #Maximum number of backing up restart file, 0 means never
    &END RESTART
  &END PRINT
&END MOTION

ultramanm87

借用Environ里面的解释：Only include/update solvation contributions when SCF accuracy is below this threshold. Since the solvation region is defined in terms of the electronic density, the test is done in order to avoid computing unphysical environment contributions, usually to skip the environ calculation during the first couple of SCF steps.


cp2k解决sccs难收敛基本上都在这儿了，你论坛里搜一下就行。
cp2k有时候sccs就是难收敛，就算某一步收敛了，结构优化后也容易不收敛。都尝试无果后建议换软件

Wenlin

ultramanm87 发表于 2025-9-25 23:29
借用Environ里面的解释：Only include/update solvation contributions when SCF accuracy is below this t ...

您好，感谢您的回复，我知道您说的SCCS部分的EPS_SCF是指当SCF迭代达到这个阈值时才激活SCCS,我是不太确定默认的0.5这个值是否正确，按理说阈值的数量级一般不都是1*10-6或更小吗？我能不能把这里的EPS的0.5改成5E-6呢

ultramanm87

Wenlin 发表于 2025-9-26 08:10
您好，感谢您的回复，我知道您说的SCCS部分的EPS_SCF是指当SCF迭代达到这个阈值时才激活SCCS,我是不太确 ...

太小或者太大都会影响收敛，一般来说我会设0.x-0.0x的级别，大体系可以让它更早开始sccs的迭代