"Engrad Freq" is put when GRRM needs energy, gradient, and Hessian.
The difference in the final Gibbs energy would be due to differences in the treatment of rotational degrees of freedom. In GRRM, rotational components that should be zero at exact stationary points are eliminated explicitly before diagonalizing the Hessian.

The TS has an open-shell singlet character. Therefore, when optimizing this TS with Gaussian, one needs to prepare initial MOs having an open-shell singlet character. In GRRM, such MO guesses are prepared automatically when the "Stable = Opt" option is given. If one prepares an appropriate initial MOs and reads them by the "guess=read" option, then one can obtain the TS obtained by GRRM even with Gaussian.

Dear all

My job seems have an error,but I can not find.Can you help me?

Dear all

I have alreadly solved,thanks

To use GRRM17 with xTB, you need to create an interface.
For details, refer to: https://afir.sci.hokudai.ac.jp/documents/manual/48

As for GRRM20, which has a built-in interface with Orca, you can use xTB via Orca.
For details, refer to: https://afir.sci.hokudai.ac.jp//documents/manual/101

Dear professor

I can not write the code to link XTB with GRRM17, dear developers, could you help me to write the code? Thanks very much 

The recommended input is as follows.

---
%link=siesta-3.1
# MC-AFIR
 
-1 3
O       0.339040   -5.411332    0.066717   1
N      -0.451510   -4.464904    0.147886   1
Ag      2.711480   -2.174275    0.122307   2
Ag      4.097417    0.239973    0.125012   2
Ag      1.473994    0.003302   -1.320331   2
Ag     -1.343819   -0.124536   -1.334256   2
Ag     -0.141218    2.402042    1.513018   2
Ag     -2.639545    2.008827    0.121617   2
Ag     -0.135432   -2.266065    0.131708   2
Ag      1.463031    0.008055    1.500499   2
Ag      2.381414    2.442301    0.117082   2
Ag     -0.139398    2.394401   -1.325776   2
Ag     -3.193985   -1.579242    0.131599   2
Ag     -1.349875   -0.119551    1.600288   2
Options
NFault = 36
Add Interaction
Fragm.1 = 3-14
Fragm.2 = 2
Fragm.3 = 1
1 2
1 3
2 3 -
GAMMA = 600
END
NOFC
SiestaProc=10
---

This calculation provides AFIR paths for the N-O dissociatoin.
Then, you need to perform the RePATH calculation to find TSs from the AFIR paths.

The recommended input for the RePATH calculation is as follows.

---
%link=siesta-3.1
%infile=xxx <== this line should be modified (xxx should be replaced accordingly)
# RePATH
 
-1 3
O       0.339040   -5.411332    0.066717   1
N      -0.451510   -4.464904    0.147886   1
Ag      2.711480   -2.174275    0.122307   2
Ag      4.097417    0.239973    0.125012   2
Ag      1.473994    0.003302   -1.320331   2
Ag     -1.343819   -0.124536   -1.334256   2
Ag     -0.141218    2.402042    1.513018   2
Ag     -2.639545    2.008827    0.121617   2
Ag     -0.135432   -2.266065    0.131708   2
Ag      1.463031    0.008055    1.500499   2
Ag      2.381414    2.442301    0.117082   2
Ag     -0.139398    2.394401   -1.325776   2
Ag     -3.193985   -1.579242    0.131599   2
Ag     -1.349875   -0.119551    1.600288   2
Options
NOFC
LUP=SingleStep
SiestaProc=10

The option Stable=Opt … END can be put into the Options section, either before or after Molpro Input … END block.

For example:

----- Sample Input -----
%link=molpro2006
# MIN

0 1
C         -0.079213112255          0.000038936045         -0.592031587948
O          0.018800714554         -0.000018308130          0.717081381477
H          0.322739833191          0.909186505065         -0.947006485140
H          0.322739833191         -0.909359858691         -0.946723086445
Options
MaxStepSize=0.1
MolMEM = 50
Molpro Input
BASIS=6-311+G(d,p)

{MCSCF,maxit=39,GRADIENT=1.d-8
 {ITERATIONS;DO,UNCOUPLE,11,TO,39;}
 START,*
 ORBITAL,*
 Occ,12;
 Closed,4;
 WF,16,1,0;
 STATE,2;}

{RS2,Mix=2,Root=2,SHIFT=0.3,MaxIt=99,MaxIti=999;
 ORBITAL,IGNORE_ERROR;
 WF,16,1,0;
 STATE,2;}

FORCE;
---
RSPT2 STATE 2.1
END
Stable=Opt
BASIS=6-31G

{HF;
 WF,16,1,2;}

{MCSCF,maxit=39,GRADIENT=1.0d-6
 {ITERATIONS;DO,UNCOUPLE,11,TO,39;}
 START,2100.2;
 ORBITAL,*
 Occ,12;
 Closed,2;
 WF,16,1,0;
 STATE,4;}
END

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

Note: This option Stable=Opt … END is only available with Molpro, and it is different from Stable = Opt option, which works only with Gaussian. For details, please see "Hints to avoid SCF convergence errors in the ADDF or AFIR searches" section in How to run GRRM with MOLPRO manual page.

Did you check the Molpro output file (XXX_MolJOB.out)? Was the Molpro calculation terminated normally? You can see the string "Variable memory released" in the bottom of the Molpro output, if the Molpro calculation was normally terminated. If you use Molpro 2015 or later, please see https://afir.sci.hokudai.ac.jp/boards/75.

How long the filename did you specify? Molpro may recognize filenames up to 31 character long (it depends on your environment), and it might be a problem...

Yes. If you run GRRM17 in multiple nodes using MPI, you must specify a non-MPI (serial) version of SIESTA executable as "subsiesta", because MPI-parallelized GRRM17p processes cannot call any "mpirun" command (current version of GRRM17 binaries does not support nested parallelism). In a parallel run, GRRM17 calls the command

${subsiesta} < XXX_SiestaJOB.fdf > XXX_SiestaJOB.out                 .

Please see the last paragraph of How to run GRRM with SIESTA.

This issue seems to be solved by the above comment. (However, you can still add comments to this thread.)

I have tested a SC-AFIR calculation using your sample input; the artificial force was not applied between any fragments because the two moieties were too far away (the input structure (EQ0) had judged as a dissociation channel (DC) structure by the default parameter DownDC=8, and you would be able to see the string "Not to be applied" in the file jobname_EQ_test.rrm.). Therefore, no reaction paths had been computed and no TS structures had also been generated.

Does the intermolecular distance have a meaning (e.g. experimentally solved structure)? If no, it might be worth to try a MC-AFIR calculation. In addition, I think "Stable=Opt" option is recommended in your case.


----- Sample Input for MC-AFIR -----

# MC-AFIR/B3LYP/6-31G
 
-1 1
C -6.04851  1.08683 -0.12952  1
C -4.65656  0.67303 -0.47198  1
C -3.56045  0.80445  0.37107  1
C -2.26684  0.40808  0.06896  1
H -6.39564  1.86874 -0.83809  1
H -4.49975  0.21350 -1.45756  1
H -3.73818  1.26439  1.35332  1
H -2.03118 -0.05733 -0.89259  1
C -7.06260 -0.07494 -0.17775  1
H -6.60275 -1.00092  0.30256  1
H -6.06390  1.54725  0.87543  1
H -1.44681  0.54598  0.77726  1
H -7.32889 -0.30211 -1.26268  1
C  3.12328  4.11991 -0.42148  2
C  1.91711  4.54010  0.00136  2
C  0.68735  4.38233 -0.75186  2
C -0.52022  4.79402 -0.32435  2
H  4.02317  4.26027  0.18117  2
H  3.24530  3.62228 -1.38871  2
H  1.82997  5.03655  0.97605  2
H  0.77488  3.88934 -1.72815  2
H -0.64143  5.28864  0.64465  2
H -1.42164  4.64671 -0.92286  2
H -8.00546  0.21979  0.39125  1
Options
NFault = 50
Add Interaction
GAMMA = 1000
Fragm.1 = 2-4
Fragm.2 = 14-17
1 2
END
Stable = Opt
GauMem = 2000
GauProc = 4
 

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

This is just my comment, and this might not be a solution to your problem...

Saita-Sensei

Thank you for the great advice!

I'm trying to shorten the distance between two fragments, to use the options, and to generate multi-structires by using MC-AFIR.

BTW... could you teach me why the fragment ranges were changed? Did the wide ranges have bad influence on the reaction?

Best regards,

S. Mieda

In the above sample input for MC-AFIR, only the sp² carbon atoms are selected for the fragment. It would be a natural choice in terms of the chemical knowledge, but it was not a strong suggestion. In order to make a global search, the wider fragment range would be required, but it would take a long computational time. So, I chose only the sp² carbon atoms to the fragments as a sample job.

NOTE: The terms "part" and "fragment" are used in different meanings in GRRM17 program:

· "Fragment" means the fragment to which the artificial force is applied by the AFIR function.
· "Part" means a moiety of the system (or a monomer in the molecular cluster). The defined parts to be used for random geometry generation (all parts to be randomly distributed).

The fragment is not necessarily identical to the part. For details, please see the manual page.
In the SC-AFIR calculation, the fragments to be automatically defined, so that you do not have to put the lines "Fragm.1 = 1-13,24" and "Fragm.2 = 14-23" into the SC-AFIR input file.

(For your information) this might be an example input file for SC-AFIR local search:

----- Sample Input for SC-AFIR -----

# SC-AFIR/B3LYP/6-31G
 
-1 1
C -6.04851  1.08683 -0.12952  1
C -4.65656  0.67303 -0.47198  1
C -3.56045  0.80445  0.37107  1
C -2.26684  0.40808  0.06896  1
H -6.39564  1.86874 -0.83809  1
H -4.49975  0.21350 -1.45756  1
H -3.73818  1.26439  1.35332  1
H -2.03118 -0.05733 -0.89259  1
C -7.06260 -0.07494 -0.17775  1
H -6.60275 -1.00092  0.30256  1
H -6.06390  1.54725  0.87543  1
H -1.44681  0.54598  0.77726  1
H -7.32889 -0.30211 -1.26268  1
C  3.12328  4.11991 -0.42148  2
C  1.91711  4.54010  0.00136  2
C  0.68735  4.38233 -0.75186  2
C -0.52022  4.79402 -0.32435  2
H  4.02317  4.26027  0.18117  2
H  3.24530  3.62228 -1.38871  2
H  1.82997  5.03655  0.97605  2
H  0.77488  3.88934 -1.72815  2
H -0.64143  5.28864  0.64465  2
H -1.42164  4.64671 -0.92286  2
H -8.00546  0.21979  0.39125  1
Options
NRUN = 1
Add Interaction
Target = 2-4,14-17
GAMMA = 200
END
SC = InterOnly
Stable = Opt
GauMem = 2000
GauProc = 4
 

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

NRUN is used in order to generate a random initial structure.
SC = InterOnly limits the number of paths to be computed (artificial force to be applied between different parts only).

This is just an example. The SC-AFIR has various options, so you may wish to experiment to decide which one fits your needs best.

Saita-sensei

Thank you very much for the useful comments.

I understand the difference between Frangments and Parts (and Targets).

Especially, SC = InterOnly option seems to be very useful for my calculation. I'm trying to use the other options for SC-AFIR, too.

If the reaction is carried out, I write the input file here to share the information.

Best regards,

S. Mieda 

This issue seems to be solved by the above comment. (However, you can still add comments to this thread.)

In your environment (on your workstation), can you execute SIESTA as a standalone program (without GRRM17) without such errors?

Please confirm your SIESTA program has been installed succesfully and you have set the environment variables "subsiesta" and "submpi" properly.

Yes, the siesta program works Ok and I have set the environment variables "subsiesta" and "submpi" properly.

Thanks in advance.

Did you put the pseudopotential files (*.psf) in the same directory? In the sample case, two psf files O.psf and H.psf are required.

This issue seems to be solved by the above comment. (However, you can still add comments to this thread.)

What options do you specify in your GRRM17 input file?

In GRRM17, the IRC calculation always computes forward and backward IRC paths (unless the "Meta-IRC” option is specified).

Of course, the forward and the backword paths can be computed from a saddle point (transition state). In other words, from a non-stationary point, only a mass-weighted steepest-descent path can be computed.

Did you see the string "IRC FOLLOWING (FORWARD) STARTING FROM FIRST-ORDER SADDLE” in your GRRM17 log file? If you got "STEEPEST-DESCENT PATH FOLLOWING STARTING FROM NON-STATIONARY POINT”, it meant your input structure was not recognized as a saddle point.

Even if your input structure was recognized as a 1st-order saddle point, it might correspond to the transition state structure between two conformers of the product.

When you use the pre-optimized transition state structure by Gaussian (without GRRM17), such a problem sometimes appears. The SCF cycles might be converged in a different electronic structure. Try using the “MO Guess = filename.chk” option in order to read the Gaussian Checkpoint file as a proper initial guess.

Or try the SADDLE calculation before the IRC calculation. The IRC paths can be automatically computed by specifying “Saddle+IRC” option.

The DS-AFIR calculation would be able to help you to get the IRC paths between the reactant and the product. (Like the QST2 method in Gaussian.)

Thank you very much for your kind reply.

In my GRRM17 input file, I specify the option IRC. I see in my file.log "STEEPEST-DESCENT PATH FOLLOWING STARTING FROM NON-STATIONARY POINT" but at the end of the file also I see: "IRC following along both forward and backward directions were finished"

I converted the file.log  in file.mol to see both movements in gaussview and, at this point, I can see only the forward movement. But as it is indicated in the file.log (IRC following along both forward and backward directions were finished), should doI see also the backward movement?

Best regards

No. The string "IRC following along both forward and backward directions were finished" just means the IRC calculation was normally terminated (termination message). Your GRRM17 output file (file.log) says that your input structure (pre-optimized TS structure obtained by Gaussian) did not converge in a saddle point in GRRM17. From a non-stationary point, only a mass-weighted steepest-descent path can be computed (the "backward" path cannot be theoretically defined).

Try SADDLE calculation with "Saddle+IRC" and “MO Guess = filename.chk” options. You will get an actual TS structure and the "forward" and "backward" IRC paths.

For example:


# SADDLE/B3LYP/6-31G**

0 1
C  -0.079213112255    0.000038936045  -0.592031587948
O   0.018800714554  -0.000018308130   0.717081381477
H   0.322739833191   0.909186505065  -0.947006485140
H   0.322739833191  -0.909359858691  -0.946723086445
Options
Saddle+IRC
GauProc = 4
GauMem = 100
MO Guess = filename.chk

This issue seems to be solved by the above comment. (However, you can still add comments to this thread.)