Error reading in TMP file 53/0 (11024)

Questions
#1

When I want to take the wave function of the small basis set as the initial guess of the big basis set,an error occured:

Guess MOs from SCF MO coefficient file
 Reading MOs from coefficient file
 Size of previous SCF MO coefficient file: 3806036 (expected); 1984030 (actual).

 Q-Chem warning in module /scratch/svnadmin/18824_zgan/qchem/gesman/GuessMan.C, line 722:

 Inconsistent size for SCF MO coefficient file.  Please check basis and PURECART setting.


 Q-Chem fatal error occurred in module /scratch/svnadmin/18824_zgan/qchem/libmdc/newfileman.C, line 293:

 Error reading in TMP file 53/0 (11024)

My input file sp_0.in as follows:

$molecule
read sp_0.xyz
$end

$rem
exchange  xygjos
basis  6-311+G(2d,2p)
aux_basis rimp2-cc-pvtz
solvent_method  pcm
dft_d   d3_bj
mem_static 600
$end

$solvent
dielectric 78.4
dielectric_infi 1.76
$end

@@@

$molecule
read
$end

$rem
exchange  xygjos
basis  6-311+G(3df,2p)
aux_basis rimp2-cc-pvtz
solvent_method  pcm
scf_guess read
dft_d   d3_bj
mem_static 600
$end

$solvent
dielectric 78.4
dielectric_infi 1.76
$end

My sp_0.xyz file:

$molecule
-4 1
 O                  1.92938953   -1.90925047   -2.78254339
 P                  2.64900286   -2.43034408   -1.54747799
 O                  2.43886233   -3.91093716   -1.27193686
 O                  4.09190754   -1.96561986   -1.42313354
 O                  1.82813112   -1.60756652   -0.30189251
 P                  1.49853568   -1.97599615    1.21026257
 O                  0.23725106   -2.78021549    1.30502853
 O                  2.70008081   -2.44844239    1.96729720
 O                  1.06974219   -0.48936506    1.76124974
 P                  1.80037300    0.92183100    2.03478500
 O                  1.44362592    1.35933985    3.41759031
 O                  3.22117063    0.91362107    1.58174668
 O                  0.92123100    1.89099000    1.06108200
 C                  1.54609562    2.56910845   -0.02492435
 H                  1.96317558    3.51887273    0.32049516
 H                  2.35361575    1.96597574   -0.43911304
 C                  0.51293488    2.83635412   -1.09786860
 H                  0.96643542    3.45372488   -1.87442493
 O                  0.09838476    1.59087133   -1.70755680
 C                 -1.26157437    1.32103809   -1.43149513
 H                 -1.87154460    1.48514001   -2.32051546
 N                 -1.43431492   -0.07220631   -1.08773660
 C                 -1.00434651   -0.73549183    0.05179249
 H                 -0.48937854   -0.22524703    0.84829559
 N                 -1.27993648   -2.01162475    0.02001783
 C                 -1.91503054   -2.21637529   -1.19328690
 C                 -2.44380531   -3.40005631   -1.78159093
 O                 -2.46783980   -4.54806883   -1.34791543
 N                 -3.00417121   -3.12047159   -3.04906453
 H                 -3.41007854   -3.91959579   -3.51600796
 C                 -3.05199930   -1.89443762   -3.65613554
 N                 -3.68073004   -1.81902861   -4.85356284
 H                 -3.82554740   -2.64930026   -5.40314818
 H                 -3.55859717   -0.96219446   -5.36740141
N                 -2.55716185   -0.80803143   -3.10753536
 C                 -2.01285757   -1.02043746   -1.89449134
 C                 -0.76752159    3.49773352   -0.59499122
 O                 -1.31104735    4.30990457   -1.62744169
 H                 -2.27120073    4.29913311   -1.51478743
 H                 -0.59474281    4.08768072    0.30704911
 C                 -1.68554267    2.29695173   -0.33012485
 H                 -1.46974354    1.88418432    0.65110956
 O                 -3.06299070    2.59945306   -0.47384162
 H                 -3.42519681    2.83572189    0.38546117
$end

Command:qchem -save -nt 16 sp_0.in sp_0.out save

PS:When I changed the molecule such as H2+,it works fine.

#2

Hello Cynthia87,

The second job uses different basis so you can’t use SCF_GUESS = READ as Q-Chem will check the consistency of size of basis. You can remove SCF_GUESS for the second job to avoid the crash. What you want is BASIS2 calculations with basis projection. (4.4.5 Basis Set Projection‣ 4.4 SCF Initial Guess ‣ Chapter 4 Self-Consistent Field Ground-State Methods ‣ Q-Chem 5.4 User’s Manual)

#4

Hello,kyliu.Thanks for your reply! I will try what you said,but when I change the system such as H2+,it works fine,Why?

#5

The reason it works for H2+ is that for hydrogen those two basis sets are exactly the same.

To make this work, either remove SCF_GUESS=READ in the 2nd job, as Kuan-Yu suggests, or else leave that alone and set BASIS2 equal to what BASIS was in the first job. That way, Q-Chem reads the previous basis as BASIS2, for which the SCF should already be converged, then projects that guess into the target basis (specified with BASIS).

#6

Hi, jherbertThanks for your reply! I tried to rerun it by setting basis2,but scf failed to converge in the second job. :face_with_thermometer:

  180   -2679.4152347059      3.52E-02
  181   -2654.7659009239      3.51E-02
  182   -2698.0019139156      3.33E-02
  183   -2695.3685875301      3.33E-02
  184   -2669.7032974580      3.40E-02
  185   -2661.2097516194      3.50E-02
  186   -2646.5751099991      3.56E-02
  187   -2645.5800414597      3.59E-02
  188   -2642.4578366675      3.62E-02
  189   -2596.3783922360      3.15E-02
  190   -2655.1499844779      3.55E-02
  191   -2548.0761993952      4.03E-02
  192   -2481.9855956726      4.37E-02
  193   -2204.1977895034      5.27E-02
  194   -2434.8497671806      4.16E-02
  195   -2494.9933621298      3.98E-02
  196   -2506.4922996427      3.99E-02
  197   -2505.5696858453      4.00E-02
  198   -2532.3802045322      3.93E-02
  199   -2573.5596389433      3.56E-02
  200   -2606.4717609587      3.59E-02    Convergence failure

 Q-Chem fatal error occurred in module /scratch/svnadmin/18824_zgan/qchem/scfman/scfman.C, line 5029:

 SCF failed to converge
#7

Not sure what’s going on, would have to look more carefully into it. But this is not an example where basis-set projection is likely to yield a lot of speed up anyway so I suggest you simply omit BASIS2 and also omit SCF_GUESS=READ

#8

It works fine to calculate the single point energy for smaller basis set 6-311+g(2d,2p),but it not for larger basis set 6-311+g(3df,2p).If it doesn’t work to do basis set projection,Maybe set the SCF_AlGORITHM=RCA_DIIS?

#9

A couple of comments:
(1) You can try changing SCF_ALGORITHM but except in specialized cases this is unlikely to work, except maybe by luck, in my experience. Stalled SCF convergence, which is what you are seeing, is almost always the result of a poor SCF guess so that is what you should be thinking about changing.
(2) Along those lines, maybe use basis-set projection from something small, say BASIS2=STO-3G (first choice) or =6-31+G* if that doesn’t work.
(3) When I run your inputs I notice a warning about the value of THRESH. You should take that warning seriously, it has to do with linear dependencies in large systems. Paradoxically, increasing the value of THRESH to 10 or 12 (or 14 in the worst case) can sometimes speed up the calculation. Slightly more time per cycle but many fewer cycles due to fewer numerical linear dependencies. In my opinion the default value of THRESH is set too loose.

#10

Thanks for your reply! When I set scf_algorithm rca_diis,it worked noramlly.I will take the warning about the value of thresh seriously

#11

When I set thresh 14，and use default diis algorithm，it worked fine.

#12

That’s great. FYI, while I consider the defaults to be too loose, that warning is sometimes a little conservative. THRESH=12 is fairly robust for large system, in my group’s experience.

We are considering a change in the default value but it needs to be considered carefully because it potentially changes a lot of behavior.