AFIRサンプル

Boards:15 Issue

New Issue

SC-AFIR option 'SC=NegativeON'
Ayako Furuhama Aug. 8, 2018, 7:25 a.m. #64

Dear Professor Maeda and his group members;

SC-AFIR calculation is performing to obtain various initial reactants of a reaction. But the following calculation is not stopped. [I do not know the reasons... Maybe 'SC=NegativeON' (?) This was used in previous calculations.] In this case, I will not follow PT structures and their TS structures due to computational cost.

Please let us know the additional or unnecessary options for such calculations. input_PARAM.rrm is following:

INPUT DATA SET OF THE GRRM VER. 13.x
-------------------------------------------------------------------------
Energy Calculation = GAUSSIAN03
-------------------------------------------------------------------------
-------------------------------------------------------------------------
Job type = 16
-------------------------------------------------------------------------
UB3LYP/6-31G*
-------------------------------------------------------------------------
0 1 XX XX 0
-------------------------------------------------------------------------
XYZ  coordinate
-------------------------------------------------------------------------
-------------------------------------------------------------------------
-------------------------------------------------------------------------
-------------------------------------------------------------------------

NRun = 32
IniStr = Random
GauMem = XXX (MW)
GauProc = X
ADD INTERACTION
SC = NegativeON
DownDC = 15
RandDC = 15
STABLE = OPT
READ BOND CONDITION

Thank you very much!

DS-AFIR issue 2
Seiji Mori July 27, 2018, 11:33 a.m. #57

Using the DS-AFIR to calculate the path that connects two structures, we found out that the calculation showed a final message  as follows:
     
The structure is dissociating
OPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTO
Path-MIN-optimization failed.....
 
If the error is caused because one of the intermediates is dissociating,  we are wondering how we can solve this issue? 
 
I tried using the  SC=FindDC  along with  %infile=NAME command  to retrieve the information obtained in previous calculations but I obtained the same error.  Is there a way to avoid the failure due to structure dissociation so the DS-AFIR is completed? 

Thank you very much in advance.
 

DS-AFIR issue
Seiji Mori July 12, 2018, 3:19 p.m. #55

 
-During a DS-AFIR calculation, AFIR program found 5 Equilibrium Structures (EQ0, EQ1, EQ2, EQ3, EQ4, where EQ4 and EQ0 are reactant and product respectively)  and 2 Transition states ,TS0 (which connects EQ0 with EQ1) and TS1 which connects EQ2 with EQ3). But, we missed  the other two TSs between EQ4- EQ3 and  between EQ2-EQ1.
 
Which option would be, in principle, more effective to find a path? To do RePATH (in the same folder where DS-AFIR was performed) or to calculate new DS-AFIRs to find the missing TSs from the each pair of EQ structures?
 
Also, another thing that caught our attention in this particular DS-AFIR output was the highlighted with a # 'hashtag' word #RIDGE-POINT (we could find this in the manual) that appears instead of "#STEP N " in the main output XXX.log . This word seems to mark the final structure in the PATH LUP   (after the structure marked with that word, the next line is this: "DS-AFIR COMPLETED, LUP-PATH OPTIMIZATION CARRIED OUT" ).  We are wondering how we  find the TS rather than the "ridge point"?
 
Thank you very much.
 
  • Re:
    Satoshi Maeda

    In general, I recommend to do RePATH in which the previous DS-AFIR results are read.

    The ridge point will be used as the TS guess only when the DS = SingleSTEP option is used. Without this option, LUP is performed after the AFIR path is obtained, and energy maxima along the final LUP path are used as the TS guesses.

1 GUI software for Gaussian outputs generated by AFIR
Seiji Mori July 6, 2018, 3:24 p.m. #53

As far as we understand, the GRRM17 uses a part of the Gaussian program (or other ab initio programs associated to it) to calculate gradient vectors and Hessian matrices. So, the XXX_GauJOB.log files (Gaussian format) generated during AFIR calculations (MC-AFIR, DS-AFIR, SC-AFIR) or minimizations contain, temporarily,  the information of the Hessian.  So our question is, what kind of other information is stored in these xxx_GauJob.log files? Are they useful as an output of the AFIR calculation?  Also, when we try to visualize these files with a output GUI visualizer like Chemcraft or Gaussview,  it shows the file as a "not-finished" optimization or a "cannot-read-properly" frequency calculation.  
In a similar manner, it is not possible to visualize the output of the GRRM17 (log files)  in a chemistry visualization program like Chemcraft, Gaussview, etc. In your paper about GRRM17 ( J. Comput. Chem.201839, 233–251https://doi.org/10.1002/jcc.25106 ), you mention a visualization code of the reaction pathways, SAFIRE, but there is no more information in the AFIR website or the paper regarding how to obtain SAFIRE. For example, if a user wants to visualize the normal modes of vibration, from the xxx_TS_list.log output file, this file is not recognized by any visualizers to show explicitly how the atoms move in their respective TSs.  Which also makes me wonder how we can visualize the result obtained in all the outputs and intermediate files (to control or keep an eye on the process of the calculation, visually) ? 
 
Thank you very much for your help.
 
 
  • Re:
    Satoshi Maeda

    We usually convert EQ_list and TS_list into the xyz format and see allthe  structures through a software which can read the xyz format.

    SAFIRE can be used to draw the network of reaction pathways. It will be uploaded to GitHub in the near future, once all required documents are available.

    Another software is Visomin (http://www.science-technology.jp/index.html). But, I have never used it by myself.

    • Thank you
      Seiji Mori

      Thank you very much for your comments.

When we are doing a energy minimization in Gaussian 09, one of the issues that rise are that the energy of minimization steps will start to oscillate and never converge until one kills the calculation job or it passes a huge number of steps, wasting a computational time. So, our question is: can this kind of behavior happen in the GRRM during the minimization performed during a AFIR calculation (or other GRRM calculation that requires the minimization algorithm) that we have to keep an eye on it to avoid wasting computer time in a calculation that gets "derailed " from its original purpose?, in other words, what are those things that we should watch out during the calculation to know that something is wrong and we have to reset or change something because the calculation won't stop but it is going wrong? 

Thank you very much.

  • Re:
    Satoshi Maeda

    GRRM uses a Quasi-Newton algorithm in geometry optimization. Therefore, common problems in Quasi-Newton algorithms also occur in GRRM calculations. In AFIR path calculations, such a case is detected and the path calculation will be terminated before wasting computational time. It is also noted that such paths are not just discarded but are processed like other paths to find TSs along the paths.

1 Determining clusters of equivalent structures
Matthew Addicoat July 4, 2018, 3:48 p.m. #48

We are using MC-AFIR to look at some reactions on metal clusters, similar to the gold cluster work here: J. Phys. Chem. C, 2015, 119 , pp 11120–11130.

We're having many equivalent structures show up as new, and while we've tried various settings of MatchDecScale and MatchDecTarget, we've not had much success in cutting down the number of incorrectly identified new structures.

Are there any other recommended options to more closely identify structures as equivalent? What options were used for the MC-AFIR of Au_n + H2? 

Many thanks,
Matt

  • Re:
    Satoshi Maeda

    MatchDecScale and MatchDecTarget do not work with MC-AFIR. In SC-AFIR, when these options are used, the GRRM program does a structure clustering and avoids the SC-AFIR searches around similar structures (those belongs to the same cluster).

    In MC-AFIR, you may try StructCheckThreshold or ScaleStructCheckThreshold. My recommendation is ScaleStructCheckThreshold = 2.0 or 3.0. Please adjust the value for your system with some tests.

4 EQ list in SC-AFIR2 calculation
Ayako Furuhama July 3, 2018, 1:21 p.m. #44

Yesterday, I attended GRRM18 tutorial and have a question.

When SC-AFIR2 is performed (input name: butadiene.com ) to consider dummy EQs, how to understand the Energy in a parenthesis.

 

For example, EQ2 is dummy EQ with "Energy    = -154.769690740354 (-154.850619271653 :    0.000000000000)", please let us know the meanings of -154.769690740354 and -154.850619271653, respectively. 

%grep Ene butadiene_EQ_list.log

Energy    = -154.864576395665 (-154.864576395665 :    0.000000000000) #EQ0

Energy    = -154.859935087313 (-154.859935087313 :    0.000000000000) #EQ1

Energy    = -154.769690740354 (-154.850619271653 :    0.000000000000) #EQ2...dummy EQ

Energy    = -154.742435832251 (-154.850514028576 :    0.000000000000) #EQ3...dummy EQ

Energy    = -154.703735106368 (-154.841020263992 :    0.000000000000) #EQ4...dummy EQ

 

Thank you in advance.

  • Sorry
    Ayako Furuhama

    If my question is inappropriate, please remove the comment. 

     

  • Re:
    Satoshi Maeda

    First, it's a tutorial of GRRM17, not GRRM18.

    For the dummy EQs obtained by the SC-AFIR2 calculations, the number before the parentheses corresponds to electronic energy plus the force term of the AFIR function. The first number in the parentheses is (bare) electronic energy.

  • Thank you
    Ayako Furuhama

    Prof. Satoshi Maeda

    Thank you very much for the clear answer. I understand the differences.

    Yes, it was a tutorial for GRRM17. A. Furuhama

  • Third energy?
    Matthew Addicoat

    And what is the third energy (zero in the excerpt above, but not always)?

    Many thanks,
    Matt

2 Input file name
Seiji Mori June 27, 2018, 6:53 p.m. #35

We understand that two files are required for the submission, for example, a shell script file (ABC.sh) which is giving the orders to the queuing system and an GRRM input file (ABC.com)  in the same folder.

After we failed one GRRM calculation with an error, and modify/overwrite the input file as the name followed by resubmission, this calculation did not work. We think that the error was caused by the presence of the residual files from the previous failed calculations. Once we deleted all the extra files and left only the ABC.sh and ABC.com,  the calculation runs successfully. Why did this error happen and it is not mentioned in the manual? Is there any possibility that the presence of other files in the same folder can interfere with the GRRM calculation?

Thank you very much for  taking time on so many questions!!

 

1 Input files for MESX or MECI calculations
Seiji Mori June 27, 2018, 6:01 p.m. #34

- In the MESX or MECI calculations, (https://afir.sci.hokudai.ac.jp/documents/manual/37) we need to define a Second Input, where we explicitly specify the level of theory and the charge and spin of the other PES that is going to cross the first one. But what we don't see in the manual is how to specify the extra Gaussian input in our second Input, i.e., if we need also to repeat the specification of the extra keywords and lines, like when you define your own basis sets or add ECPs by using a GenECP keyword to the Second Input? or we do not need to repeat that and rather specify in the first input file (the ones defines in the previous GauInpB) ?

https://afir.sci.hokudai.ac.jp/documents/manual/20

 

Thank you very much again.

 

  • Re:
    Satoshi Maeda

    In OptX calculations, the common GauInpB is used  for both of the two states.

2 Meta-IRC issue
Seiji Mori June 27, 2018, 5:52 p.m. #33

-If we want to calculate the IRC trajectory of a structure, it is possible to do it with GRRM with the keyword  IRC, and it will give you the IRC path in both directions, the forward and the backwards direction to the Gradient Vector. Now, if you want only one of the directions of the IRC path, you could use the keyword Meta-IRC to compute ONLY the forward path. is there a way to compute ONLY the backward path?  I am assuming the forward and backward concept from Gaussian, in which they refer to the direction of the gradient Vector, which means that sometimes the Product path can be in the Reverse direction and not the Forward direction. 

https://afir.sci.hokudai.ac.jp/documents/manual/14

  • Re:
    Satoshi Maeda

    The meta-IRC is the steepest descent path in the mass-weighted coordinates. Therefore, it leads toward the inverse direction of the gradient vector (even when the norm of gradient vector is very small). 

  • If that is the case, that the meta IRC calculates the backward path (inverse direction) how can we calculate the ONLY the forward direction? because the IRC keyword gives me both sides, Forward and backwards, not only one of them. Thank you.