Maximum number of SCF cycles reached for frequency calculation

Questions
#1

Hello,

I am trying to perform a frequency calculation on an optimized geometry that we have some evidence for experimentally. I have set a large number for the maximum number of SCF cycles (6000) and have SCF_CONVERGENCE set to 8. SCF fails to converge after the 6000 cycles. I was wondering if there is something I have set wrong or if there is some way I could resolve this problem. I can’t imagine that I should have to set more than 6000 maximum cycles. My input file is pasted below.

$molecule
1 2
  C   -2.0983822   -0.7707572   -1.7318422
  N   -1.3151515   -1.7528459   -1.2353028
  C   -1.8212724   -3.0021230   -1.1840898
  C   -3.0995405   -3.3078448   -1.6272999
  C   -3.9193322   -2.2998724   -2.1502693
  C   -3.3840734   -1.0030176   -2.1888689
 Ru    0.6657386   -1.4743375   -0.6089945
  O    0.3078988   -2.7710100    0.9562896
  C   -0.0341560   -2.1921341    2.0806848
  O   -0.2470690   -2.7756882    3.1263991
  C    3.3797849   -0.2209490   -0.0499302
  C    4.7187648   -0.1911510    0.2998138
  C    5.3662524   -1.3730726    0.6935702
  C    4.6018106   -2.5474704    0.7024371
  C    3.2660649   -2.5051833    0.3358034
  O    1.2184058   -2.9961802   -1.8971795
  N    0.1227549   -0.2023928    0.7815144
  C    0.0507856    1.0986623    0.4665263
  C   -0.3350457    2.0557274    1.4130505
  C   -0.6377624    1.5694917    2.7109345
  C   -0.5549016    0.2107392    3.0021523
  C   -0.1647661   -0.6745556    1.9874916
  C    0.3100945    2.7641709   -1.2860203
  C    0.3876316    1.4209930   -0.8900257
  N    0.7570961    0.3861823   -1.7147649
  C    1.0571339    0.7031957   -2.9695176
  C    0.9989094    2.0170822   -3.4688928
  C    0.6259879    3.0515562   -2.6324322
  H    0.5785586    4.0807116   -3.0007919
  H    1.2589433    2.1919554   -4.5142818
  H   -0.9443880    2.2769918    3.4874503
  H   -0.7905647   -0.1769166    3.9961508
  H   -1.6877432    0.2401131   -1.7665807
  H   -3.9681987   -0.1645644   -2.5784984
  H   -1.1805693   -3.7852145   -0.7741706
  H    2.6575931   -3.4128054    0.3242981
  H    5.0457263   -3.5058024    0.9854968
  H    5.2579144    0.7593726    0.2612180
  H    2.8723366    0.6964244   -0.3587422
  O    0.6203898   -1.6672680   -4.0381230
  S    1.6638352   -0.6059418   -4.0629478
  O    2.9351134   -1.0061209   -3.4155714
  O    1.8216364    0.0233332   -5.3787360
  H   -3.4515729   -4.3408061   -1.5595783
  C   -5.3011567   -2.5833012   -2.6608479
  H   -5.6236015   -3.6030526   -2.4056033
  H   -6.0224431   -1.8603673   -2.2480815
  H   -5.3218585   -2.4749190   -3.7584390
  C    6.8082794   -1.3683924    1.1063931
  H    7.2547693   -2.3692728    1.0138446
  H    7.3878578   -0.6497965    0.5071504
  H    6.8865094   -1.0572199    2.1625770
  N    2.6551483   -1.3598364   -0.0295499
  H    0.9353390   -2.8556215   -2.8319878
  H    1.0403985   -3.9138589   -1.5383743
  C   -0.3923454    3.4324212    0.9766778
  C   -0.0862282    3.7608848   -0.3102290
  H    1.6858263   -5.5690078   -0.2787552
  O    0.8323010   -5.1408008   -0.4465933
  H    0.6769609   -4.5850018    0.3426959
  H    3.7160396   -2.5993355   -2.9792321
  O    4.0450121   -3.4423989   -2.6131237
  H    3.2301555   -3.8091094   -2.2403937
  H   -0.6890565    4.2009788    1.6958079
  H   -0.1336133    4.8020727   -0.6423113
$end

$rem
   BASIS  =  def2-SVP
   ECP_FIT  =  TRUE
   UNRESTRICTED  =  TRUE
   JOB_TYPE  =  FREQ
   METHOD  =  wB97XV
   GUI  =  0
   SCF_CONVERGENCE  =  8
   SOLVENT_METHOD  =  SMD
   MAX_SCF_CYCLES  =  6000
$end

$smx
   SOLVENT  water
$end

Thanks in advance!
Jake

#2

Transition metals can be a challenge to converge and you often need to change both SCF_GUESS and SCF_ALGORITHM. For the former, I suggest the FRAGMO guess. It looks like you have 3 ligands plus 3 waters, so make each ligand and each water a separate fragment, with the Ru as a 1-atom fragment. This allows you to set the charge and multiplicity on the metal center in your guess. For the algorithm, I have been having decent luck with ADIIS_DIIS, occasionally changing THRESH_ADIIS_SWITCH from 3 to 4 and increasing MAX_ADIIS_CYCLES. I think the guess is the more important factor, because by default you are getting a SAD guess with a charge-neutral Ru, which is not what you want.