AFIRサンプル

  1. Home
  2. Forum

Boards:19 Issue

New Issue

1 Elementary questions
YU Kaneko June 28, 2019, 9:55 a.m. #76

Dear GRRM developers,

Very recently, I became a new user of GRRM17. I managed to perform MC-AFIR calculation for exploring new reaction between different two molecules.  I have three question now.

Question 1 :  I would like to understand the meaning of "Fragment" and "Part". In "MC-AFIR", "part" means the index of molecules for generating initial configration of geometry. I found same question and answer (  Applying artificial force in SC-AFIR , https://afir.sci.hokudai.ac.jp/documents/manual/32) "Fragment" means the index of atoms that is applied to AFIR external force during reaction searching.   If so, if I could guess the atoms of reaction area properly, the whole calculation time could be reduced.

Question 2: What is the usage of "GRRM17.out" ?  If it is something for summarizing reaction pathway, I would like to use it.

Question 3: When I could not found the desired or supposed reaction pathway among the extracted reaction pathways, could  large value setting for NFault or GAMMA be effective ? In my understanding, NFault just limits the number of reaction patth collection, and GAMMA limits the walkable energy hieght during TS survey. If you have effective option,  I would like to know.

Best regards,

Yu Kaneko 

  • Sorry
    YU Kaneko

     I forgot erase the last lines... If someone could, please erase it.

GRRM linked with MOLPRO2015
Yasuhiro Shigemitsu June 25, 2019, 11:45 a.m. #75

Dear GRRM developers,

I'm trying to use GRRM17 combined with MOLPRO2015 but the combination does not work.  The GRRM tutorial reads that the version 2006, 2008,2010,2012 were successfully tested with GRRM. Kind helps or information would appreciated to use the combination, GRRM17 plus MOLPRO2015.

Best regards,

Yasuhiro Shigemitsu

  • GRRM17 linked with Molpro 2015 or later
    Kenichiro Saita

    GRRM17 does not recognize the keyword "%link=molpro2015" or later.  Consequently, GRRM17 is officially compatible with Molpro 2006, 2008, 2010, and 2012. However, you can use Molpro 2015 or later at your own risk.

     

    With Molpro 2015: 

    Whereas we have not fully tested Molpro2015 jobs, the formats of output files of Molpro 2015 seems to be the same as those of Molpro2012, and GRRM17 will work with Molpro2015 by using the link keyword "%link=molpro2012". 

     

    With Molpro 2018 or later: 

    GRRM17 knows a Molpro job has been terminated normally if the string "Variable memory released" is appeared in the Molpro output file (XXX_MolJOB.out), because Molpro2015 or earlier prints the string in the last line of the output. But, Molpro2018 or later uses "Molpro calculation terminated" instead of "Variable memory released". Thus, GRRM17 will never understand the termination of the Molpro2018 (or later) calculation, and then GRRM17 program will stop with an error.  This issue will be evaded by using the TEXT card in Molpro input. For example: 

    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;
    IF(STATUS.LT.0) STOP
    $STR='Variable memory'
    TEXT,$STR released
    ---
    RSPT2 STATE 2.1
    END

    We are aware of another issue; the format of the punch file (XXX_moljob.pun) is different from that of Molpro2015 or earlier, then GRRM17 cannot read the energies from the punch file. This issue will be evaded by setting "submol" command to the shellscript in which the punch file format to be converted. For example: 

       setenv submol "./script.csh"   (csh/tcsh case)

    and prepare "script.csh" like as 

    #!/bin/csh -f
    set PUN=`echo $2:r.pun | tr \[A-Z\] \[a-z\]`
    set TMP='tmp'

    molpro -W./ -d./ $1 $2

    grep "Energy" ${PUN} > ${TMP}
    set i = 1
    set n = `wc -l ${TMP} | awk '{print $1}'`
    while ( $i <= $n )
       set line = `head -n $i ${TMP} | tail -n 1`
       echo $line >> ${PUN}
       @ i = $i + 1
    end
    \rm -f ${TMP}
    unset i, n, line
    unset PUN, TMP

    exit

    Alternatively, you can use the general interface in GRRM17. Please see "General interface with external ab initio programs".

    We are aware of another issue; the format of the punch file (XXX_moljob.pun) is bit different from that of Molpro2015 or earlier, then GRRM17 cannot read the energies from the punch file. For example:

    RSPT2 STATE 1.1 Energy    -114.21130105   (Molpro2015 or earlier)

    RSPT2 STATE  1.1 Energy    -114.21130105   (Molpro2018 or later)

    So, make sure whether you provide an appropriate string to the Molpro Input section. For example,

    Molpro Input
    ︙
    IF(STATUS.LT.0) STOP
    $STR='Variable memory'
    TEXT,$STR released
    ---
    RSPT2 STATE  2.1
    END

     

    We have made several tests using Molpro 2015.1 (patch level 33), Molpro 2018.2, and Molpro 2019.1.2, but we cannot give any warranty.

ONIOM on MC-AFIR
Taiki Kato Nov. 27, 2018, 10:40 a.m. #71

Dear Prof. Maeda & laboratory people,
 
I'm trying ONIOM method on MC-AFIR.
But, I don't understand the exact format of the ONIOM layer & MC-AFIR part.
This is my input file with error calculation log.

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

# MC-AFIR/ONIOM(uwb97xd/6-31G : upm6) integral(coarsegrid)
 
0 1
H  -5.09198232  1.46148646  0.64415956 H   1
H  -5.05103711  1.51846702 -1.80580446 H   1
H  -5.08619301  3.61195144 -0.53172902 H   1
H  -3.07552771  2.21038282 -0.53072587 H   1
Si -4.57617575  2.20057460 -0.55601978 H   1
Si  0.92262713 -0.42951851 -0.06586408 H   2
H   0.32586471  0.89377867  0.44846657 H   2
H   0.92141623 -0.42839950 -1.60585939 H   2
Si -0.36596427 -2.22060001  0.71381911 H   2
Si  0.54148635 -4.23086015 -0.06762047 H   2

...

Si  7.89452207  2.71361008 -0.06626650 L H 16 2
Si  6.60446476  0.92311796  3.05275285 L H 16 2
Si  3.11867252 -0.64788568  3.05254379 L H 12 2
Si -2.56235026 -2.00324774 -0.06393534 L H  9 2
Si  1.83058292 -2.43960890  3.83181251 L H 19 2
Si  1.45021923 -6.24149307  3.83086153 L H 19 2
Si -0.74565359 -6.02268238  0.71234080 L H 10 2
Si  2.35719613 -8.25146150  3.04887994 L H 23 2
Si  4.55185684 -8.47158271  3.82932265 L H 33 2
External Atoms
Si  7.50985596 -1.08833081  6.95223274 L   2
Si  7.12790465 -4.89067932  6.95082004 L   2
Si  8.03487102 -6.90152356  6.17032595 L   2
Si  8.41640895 -3.09947186  6.17176893 L   2

...

H   6.74535629 -8.69308650  8.48930406 L   2
H   7.50965005 -1.08890581  8.49222140 L   2
H   4.02610987 -2.65912846  8.49169783 L   2
H   3.64455597 -6.46260745  8.48999760 L   2
H   0.54408428 -4.23167248  8.49284240 L   2
Options
MicroIt = (MMOnly)
GauProc = 4
NSample = 20
MinFC = -1
MC = ReactivePathOnly
Add Interaction
Fragm.1 = 1-5
Fragm.2 = 6-43
Fragm.3 = 66-145
1 2
1 3 -
GAMMA = 600
END

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

I think that the number order of ONIOM layer or MC-AFIR part is not good.
Please teach me the exact number order of ONIOM layer & MC-AFIR part.
 
Best regards

  • ONIOM on MC-AFIR
    Satoshi Maeda

    To combine ONIOM and MC-AFIR, please use the following format.

    Atom X Y Z  part-number Layer  (Link atom info)

    Atom:atom type
    X:X-coordinate
    Y:Y-coordinate
    Z:Z-coordinate
    part-number:number for MC-AFIR
    Layer: High/Midium/Low
    (Link atom info):as needed

1 Metal Element Containing Surface Reaction Analysis
Taiki Kato Nov. 8, 2018, 5:30 p.m. #67

Dear Prof. Maeda & Laboratory People
 
I'm struggling to analyze metal element containing surface reaction analysis such as { TiCl4 on TiN }, { EtOH on HfO2 }, { SiH4 on Ru } and so on.
 
I use GRRM17 which is not feasible PBC surface model.
Then, I use cluster surface model on these metal element containing surfaces.
However, it is difficult for me to converge these surfaces { TiN, HfO2, Ru } SCF calculation by Gaussian.
I also use Materials Studio DMol3 which is easy to converge SCF calculation but not suitable for GRRM.
 
Please teach me how to converge these SCF calculation by Gaussian or some alternative convergent technology.
 
Sencerely

  • Re:
    Satoshi Maeda

    Although GRRM17 doesn't contain an interface with DMol3, GRRM17 can be used together with any energy computation code if a user prepare a simple interface script. Please see the following page:

    https://afir.sci.hokudai.ac.jp/documents/manual/48

    • Other QM code trial
      Taiki Kato

      Thank you for your kind reply.

      I'll consider the GRRM application trial not only on Gaussian but also on the other QM codes.

      DMol3 doesn't have any analytical hessian calculation method.

      Therefore, I think that DMol3 may be very slow on GRRM.

      And I hope to choice more established way because I'm a biginner on these operations.

      How is the performance of SIESTA (or Turbomole) on cluster model analysis ?

      Will you please teach me SIESTA (or Turbomole) setting manuals ?

      If you know good QM-codes, please give me some advice.

1 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!

  • Option for SC-AFIR
    Ayako Furuhama

    Dear Prof. Maeda,

    I am sorry but I clarify my question.

    I search TS and EQ structures for the initial reaction of A + B system by using SC-AFIR/UB3LYP/6-31G(p).
    A is a tri-atomic molecule and B is an organic molecule including one nitrogen and some carbon and hydrogen atoms.
    After one month starting a SC-AFIR calculation (B= one nitrogen + three carbons with some hydorgens) with options (i.e. NRUN=32,GauMem=1000,GauProc=4,Add Interaction,and Gamma=100 alog with SC=NegativeON,SC=LUP-ON,NoBondRearrange,DownDC=15,RandDC=15,Stable=OPT), 343EQ and 629TS (as well as 5900 TS in PT_list) were found. But this job does not done yet. EQ and TS searches are not stopping.
    Could you please inform how to finish the SC-AFIR calculation properly? Should I perporm FirstOnly calculations in advance to found out proper initial conformation?

    Thank you very much in advance.
    Ayako Furuhama

     

    • Re:
      Satoshi Maeda

      I don't know your system well, but I guess that many similar (but slightly different) EQs are generated in the flat regions of the PES in your calculation. This often happens when a spin unrestricted DFT method is used as an energy calculation method.

      One can avoid this situation by one of the following three ways.

      1. By increasing the grid density in DFT calculations.
      E.g.) UB3LYP/6-31G(d) Int(Grid=99590)

      This would be the most straightforward way, but DFT calculations with dense grids will be computational demanding.

      2. By using a structural clustering option.
      E.g.) MatchDecScale = 3.0

      See the option list about this option: https://afir.sci.hokudai.ac.jp/documents/manual/51

      3. By applying a weak force between all atom pairs.
      E.g.) A simple case for CO+OH is shown below:

      # SC-AFIR/UB3LYP/D95V

      0 2
      C 0.0 0.0 0.0 1
      O 1.2 0.0 0.0 1
      O 3.0 0.0 0.0 2
      H 4.0 0.0 0.0 2
      Options
      Add Interaction
      Fragm.1=1
      Fragm.2=2
      Fragm.3=3
      Fragm.4=4
      1 2 3.33
      1 3 3.33
      1 4 3.33
      2 3 3.33
      2 4 3.33
      3 4 3.33
      Gamma=200.0
      END
      EQOnly
      KeepLUPPATH

      With the above input, a weak force with Gamma=3.33 kJ/mol is added to all atom pairs during the SC-AFIR calculation with Gamma=200.0 kJ/mol. A recommended value of the weak force is 10.0/(N-1) kJ/mol, where N is the number of atoms in the system. All EQs and TSs obtained are those on the AFIR function consisting of these force terms and thus are not actual ones. Therefore, after the initial SC-AFIR calculation, one needs to perform a RePATH calculation reading the results of the initial SC-AFIR calculation to eliminate the weak force. Since TSs obtained by the initial SC-AFIR calculation are not actual TSs, it is recommended to avoid computing TSs accurately during the initial search using the EQOnly and KeepLUPPATH options (also see ReadBareEnergy option).

      When the weak force is added, local minima for weak complexes between two molecules are formed in regions where the two molecules have sufficient interactions. When DFT grids that are not dense enough are used in spin unrestricted DFT calculations, many imaginably local minima can be created in the flat regions of the PES where two molecules are arranged far away from each other. The weak force help avoids this problem. I emphasize again that the weak force must be eliminated afterward by the additional RePATH calculation.

      An example input for the RePATH calculation is also shown below:

      %infile=xxx
      # RePATH/UB3LYP/D95V

      0 2
      C 0.0 0.0 0.0 1
      O 1.2 0.0 0.0 1
      O 3.0 0.0 0.0 2
      H 4.0 0.0 0.0 2

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., 2018, 39, 233–251,  https://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.

1 Energy oscillating issue during geometry optimization
Seiji Mori July 6, 2018, 9:52 a.m. #52

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.

↑