It seems your system is fairly large for QM. For these type of calculations it is better to use QM/MM methods in Gaussian. So, please try ONIOM method in Gaussian.

Also you can try resubmitting the failed calculations with slightly different geometries.

Hello,

You can try two different options, the first one (appropiated if the deltaE in the SCF iterations is decresing continously) is to add in the input section the keyword SCF(maxcyc=number) with this number bigger than 129. The other option is to try a quadratic convergence of the wavefunction using the keyword SCF(XQC).

Regards,

Ignacio

Thank you both for your advices!  Let's try it. :)

Hello, you can also try with the keyword "nosymm" to denie the symmetry option.

Good luck,

Quang

Thanks! :)

The problem of convergence is always complex and may be both in model or in method. The first question is the geometry - where do you have it from? It may be far from minimum (too distorted comparing to the one representing the optimized structure). If so - verify and revise the geometry. 

The other thing is the method. Your system is quite large for QM and some tweaks may be needed. At first try to tune the optimization method (algorythm, max atom translation, convergence criterion).

http://www.gaussian.com/g_tech/g_ur/k_opt.htm

You may consider performing a preoptimization first, maybe even using a non-hybrid functional (e.g. PBE) to get a good starting geometry for the fine optimization.

Good luck!

I totally agree with Daniel, however, I don't think PBEPBE is a good starting point. Because in my experience, PBEPBE causes many problem in convergence, especially with radical fragments. Maybe a semi-empirical like PM6 is better. You should pre-optimize the structure with PM6, and then fully re-optimize with your B3LYP. If not, ONIOM combined B3LYP for high layer and PM6 for lower layer is also a very choice.

Best,

Quang

Hello Alzbeta

Suggest you to  try diffuse function over  6-31G* in case of negativelly charged fragments. Do it  after relaxing the structure using MM and/or pre-optimization as suggested by previous answer. Good luck.  

The first step I would take is to increase the number of SCF cycles with the keyword:

SCF=(maxcycles=XXX)

A good number to use would be around 1024 or so. If you need more than 2048, there is probably a different problem. I just chose those numbers because they are nice powers of 2, by the way.

If that doesn't work, you could use quadratic convergence with SCF=QC

I assume you are just performing a single-point calculation on the protein because, as others have said, trying to optimize a protein at that level could take months (or at least weeks) of computing time even with a pretty good starting geometry...

Another strategy is ONIOM, which would make an optimization feasible. I see that I have already been beaten to the punch, but in case you don't know what that is, you would be treating the active/interesting residues as the high "layer" and the rest of the protein as the low layer, probably using MM methods like UFF (but maybe better). If you want to get better results, you could also treat nearby residues as the medium layer using 3-21G* or something. You can select the atoms for each layer manually in Gaussview.

Sorry for not really adding anything new, but I hope it helps either way.

Use scf=xqc (as somebody suggested earlier) and int=nofofcou.  If this does not work change the funtional. I had a lot of convergence problems with B3LYP. If you do this, begin for calculating the systems that did not converge with B3LYP. Regards, Juan

There are few things that you can try to resolve this problem:

1. increasing the number of scf cycles; scf=maxcycle = 1000

2. using quadratically convergent method; scf = qc

3. refining your starting structure (initial geometry)

4. trying the optimization at HF level, then using optimized structure as initial guess for B3LYP...

Hope these help!

Try to use denser integration grid: integration=ultrafine

Using the options: SCF=(NoVarAcc, NoIncFock) may also help.

Good Luck

By the way: geomtry far from optimum may also cause problems with the SCF convergence. If this were the case, you may also try to improve geomtry by running a few optimization steps with very small basis set (even STO-3G might help)

dear alzabeta,

see if ur working with 41 fragments according to me different atoms require different basis sets as per their atomic number .

so try

b3lyp/6-311dp ++

b3p631dp ++

b3p6311dp  ++

pbe631dp ++

pbe6311dp ++

hope it helps to solve ur problem .

You can try scf=(maxcycle=3000) scf=(conver=4). It may helpful for you.

I have faced the same problems with a Ru complex. After trying to change the algorithm (to qc, xqc... )and also starting from a lower level of theory, I would like to point it out that the only thing that really worked it was the suggestion by Srimanta Pakhira, increasing the number of cycles to 3000.

increase the number of cycles to 3000

I will assume you have an even number of electrons because G09 did not complain when you invoked RB3LYP.  I think you should try something similar to Marcin Andrzejak's suggestion.  Optimize first with PM3, then B3LYP/STO-3G, and finally B3LYP/6-31G* but with the guess=read geom=check options to read the previous calculation's solution and geometry from the checkpoint file.

In earlier quantum chemistry software a way to overcome convergence problems was used. It consisted in starting a calculation but using the average of the density matrices of the N and N+1 earlier steps. Does anyone knows if this option is inside G03 and/or G09? Txs.

Hello,

Could you please try to use max SCF cycle?

Here is the solution

You can try scf=(maxcycle=3000) scf=(conver=4). It may helpful for you.

My best

I have recently had success using the RFO algorithm for minima (not just transition states) when the optimization perpetually oscillates between two different structures. You can request it with opt=RFO. It probably only helps convergence issues if you are already sufficiently close to the minimum.

Also, SCF=(conver=4) can get you to a place on the PES that you don't want to be unless you really know what you're doing (novices beware). I'd use 5 or 6 instead if you need to resort to that.

hi alzbetha,

r u still dealing with convergence problem if yes then do let me know what r the elements involved in ur complex . becoz the choice of basis set affects the optimisation result . if it ok with u, u can send me its structure via message and i can see what basis set will work for ur complex .

hope i i hav solved ur problem.

Hi, Even after using the keyword scf=(maxcycle=2000), the calculation was terminated with an error L117.exe(Maximum Number of Iterations exceeded). What is the keyword to increase the number of Iterations in Gaussian 09?

Hi all,

i am using a 162 atom molecule for finding HOMO-LUMO levels through DFT.  i tried with different basis sets, but most of the times, i got same problem of convergence failure issue.  

So, after seeing the above comments, i tried Ambrish K. Srivastava · option of first trying Optimization with HF, but did not add any additional cycles option.  again conv. failure.

So, if i use maxcycle option, where exactly i should write the SCF=(maxcycle=3000) SCF=(conver=4).

Awaiting your reply please.

Please try running this calculation as a triplet spin state, with UB3LYP (or just B3LYP, but not RB3LYP in the route statement). You clearly have even number of electrons but your singlet calculation clearly is not leading to a good solution.

Hi Ramachandran,

Thanks for the reply.

Just now i submitted the job with 'opt=CalcFC', SCF=(maxcycle=3000) SCF=(conver=4). I am new to this software.

there was an input 'geom=connectivity polar' automatically added when saved from Gauessview.  But when i am adding the above SCF statements the line is not enough for the editor screen and gone next line. not sure how to continue in second line, i deleted this above "geom=.....' input.

will there be any problem?

i will use your suggested options again in another job.

As you may have noticed already, the overflow probably did not cause any problems. Gaussian will read the route statement from the "#" until the blank line prior to the job title.

Hi 

i am trying the following in ***.gjf input file

%chk=pv29-30Oct.chk

%mem=30GB

%nprocshared=16

# opt=calcFc ub3lyp/6-31+g(d,p) nosymm maxdisk=30GB geom=connectivity

# scf=maxcycle=1000

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

the script file has.......

#!/bin/bash

#PBS -l walltime=10:00:00

#PBS -l ncpus=16

#PBS -l mem=100GB

#PBS -l jobfs=100GB

#PBS -l software=g09

#PBS -l wd

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

after submitting the job it says :

Not running: insufficient amount of resource: job_tags

--------

Whats it all about. please?

Please change your route statement to:

# opt=calcFc ub3lyp/6-31+g(d,p) nosymm maxdisk=30GB geom=connectivity scf=maxcycle=1000

That is, just one line should start with #.  Also, you are asking for 100 Gb of shared memory in the PBS script when the input file only asks for 30 Gb. Your machine may not have 100Gb available on a single node.

Thank you Ram.

incidentally, before i try to edit, the status shows the previous job started. i think the machine is waiting for resource availability and started now. Let me see what will be outcome and then come back.

Regards.

OK. The maxdisk=30Gb specification may cause problems if your scratch files exceed that size. You don't really need that in the route statement.

hi

it did not work.

it says finished as status but in .log file no such message, but says "Restarting incremental Fock formation."

this message repeated 16 times.

'qstat command says time use 160:44:4". so much time??

not sure what to do.. :((

This is a very large molecule. It looks like you ran out of time. Why not try a PM3 calculation to generate a decent initial guess?

# opt pm3 geom=connectivity scf=maxcycle=1000

Thanks for suggestion. i just submitted. will update the outcome.

regards.

Hi

it run better than before. but still has issues. the end statements in log file, gave some idea, there might be some issues with few atom  bond angles, i suppose.  modified the structure a bit at these atoms and re submitted. The log file statements are copied.

----

Berny optimization.

Using GEDIIS/GDIIS optimizer.

Bend failed for angle 21 - 22 - 25

Tors failed for dihedral 20 - 21 - 22 - 25

Tors failed for dihedral 26 - 21 - 22 - 25

Tors failed for dihedral 27 - 21 - 22 - 25

FormBX had a problem.

Error termination via Lnk1e in /apps/gaussian/g09e01/g09/l103.exe at Tue Nov 1 02:52:21 2016.

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

hi

its finished with normal termination. but cannot open .chk file.

error.

CConnectionGFCHK::Readfile()

Cannot find file.:

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

Job cpu time: 0 days 2 hours 15 minutes 32.5 seconds.

File lengths (MBytes): RWF= 101 Int= 0 D2E= 0 Chk= 30 Scr= 1

Normal termination of Gaussian 09 at Tue Nov 1 15:42:46 2016.

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

Hi

run with CAM-B3LYP and other options.  and now it run. Though log file final statement not says "normal termination', i could open the .chk file.  The HOMO-LUMO orbitals are shown. But not matching with literature MO diagram for HOMO-LUMO. By the way the energies shown at HOMO-LUMO what are the units? and if not eV, how to convert to eV? 

Hi Saravanan,

Thank you for the reply.

1. Yes i tried with higher %mem=30 GB

2. I have only Gaussview.  is that Avagadro better? i do not have it though. Do you mean i need to develop the structure from fresh in Avagadro?

3. as my molecule is very big having 162 atoms, and a recent work on such molecule suggests CAM-B3LYP with 6-31G+(d,p).  but anyway i try your suggestion as well for optimization.

By the way, can anyone tell "How do we confirm that our final optimised structure is correct or not?

Regards

hi 

By the way i have CHEM-3D Draw and i remember i initially done the structure in it and then moved to Gaussview for creating .gjf file.  And later one done simple reorientation etc but not redrawn in Gaussview.  Does this could have created a problem at initial structure level?

Regards

Dear SKM 

i think there will no problem in structure level. you can proceed with it

Hi,

You can simply add this keyword to your in put file, SCF=QC but, it makes your calculations longer that normal times so, make sure you take this keyword off from your input after some cycles are passed. The convergence failure is usually happen when you have un-pair number of electrons such as transition metals.

Hope this help,

Hello Friends even me I am facing the same problem but in case im dealing with the transition metal complexes. So can you suggest anything I have tried all these above steps but does not work .. 

Hi Fitzerald, Can you provide us more information about your system and your error? depending on the metal and on the structure, the convergence could be a problem, especially if different spin states are close in energy... If XQC does not solve your problem, probably our geometry input is far from the minimum. Also, if you are trying to optimize a low spin state, sometimes you can optimize your system in high spin, and the read the orbitals to compute the low spin!

My system is a binuclear type of structure which have at least two transition metals. The same thing after running 129 cycles show convergence failure. I have tried with scf=xqc and scf=(qc,direct,maxcycle=512) show the same results. Below is the error at the time of convergence.

SCF Done: E(UM06L) = -1008.66426315 A.U. after 129 cycles

Convg = 0.6355D-04 -V/T = 2.1258

= 0.0000 = 0.0000 = 0.0000 = 0.0000 S= 0.0000

= 0.000000000000E+00

Annihilation of the first spin contaminant:

S**2 before annihilation 0.0000, after 0.0000

Convergence failure -- run terminated.

Error termination via Lnk1e in /usr/local/g09/l502.exe at Sat Jun 10 12:50:21 2017.

Job cpu time: 0 days 1 hours 4 minutes 23.6 seconds.

File lengths (MBytes): RWF= 83 Int= 0 D2E= 0 Chk= 3 Scr= 1 

 hi fitzerlnad,

ur working on transition metal right,  can u specify the metal being used n what basis set u hav used. sometimes the basis set being used also creates error. also tell me what is the no of atoms in ur complex along with the memory ur using . i hope ur querie is solved.

thank u

I am Running an AM1 Single point Energy calculation in order to calculate the exact charge of a metal ion located in the middle of a protein binding site as a co-factor. Same error was appeared while running the calculation. please tell me is there anything to be done specially

Input File

%NProcShared=7

#n RHF/3-21G Opt

HF calculation

1 1

tail of the log file 

Rare condition: small coef for last iteration: 0.000D+00

Rare condition: small coef for last iteration: 0.000D+00

Rare condition: small coef for last iteration: 0.000D+00

Rare condition: small coef for last iteration: 0.000D+00

Rare condition: small coef for last iteration: 0.000D+00

>>>>>>>>>> Convergence criterion not met.

SCF Done: E(RAM1) = 0.103850808928 A.U. after 129 cycles

NFock=128 Conv=0.35D-03 -V/T= 1.0001

Convergence failure -- run terminated.

Error termination via Lnk1e in /opt/g16/l502.exe at Mon Aug 14 11:06:33 2017.

Job cpu time: 0 days 0 hours 1 minutes 38.3 seconds.

Elapsed time: 0 days 0 hours 0 minutes 13.3 seconds.

File lengths (MBytes): RWF= 78 Int= 0 D2E= 0 Chk= 5 Scr= 1

Hi Chamikara,

If you are running single point energy calculation, please remove 'opt' keyword from  the routine section.

Hope this helps.

Thanks

Hi all.

I am facing the same problem for transition metal complex.Link was  died automatically when I used semi empirical method.If it shows error in semi empirical method,will it be continue to other methods like DFT, HF etc?? what is the right way to overcome this problem? As I run the program,it shows "error in internal co ordinates".

Not necessarily. Semi empirical methods have many limitations. DFT should work. For  "error in internal co ordinates", you should change/modify your starting structure. Hope this helps!

read ur querie try dft with pbepbe lanl2dz and pbepbe lanl2mb functional n basis set try opt . can u specify what atoms r being used in ur complex so that i can suggest more appropriate basis set.

hope it helps

for transition metal complex i suggest lanl2dz and lanl2mb combos with an functional they work fine with transition metals .

This system is much too big to begin with b3lyp - first converge a HF wavefunction with a small basis set - then use those orbitals as the input guess (guess=read) for a larger basis set and, eventually a b3lyp calculation - I'd do HF/3-21G first and see -

At the stage of convergence not reached, extract the coordinates at that stage of iteration before the error starts appearing. Submit this coordinate set as starting set, reduce the stringency of convergence criterion, and get a result without error. Start with this result the iterative process and increase the stringency of convergence criterion and repeat. Some times for certain type of systems, there is a progress to get good convergence.

Hi all

i think the suggestion by Sanarampadi Aravamudhan is good. Though i have not tried it. As now i temporarily stopped orking on my previous project.

re-optimize the fragment with opt=calcfc.

Dear Herath!

From the log file you attached, it seems as if your initial structure is far from the optimized one. And because there is not much information about your system, so after going through your error file, I suggest you to try two things:

1. take somewhat better guess for structure.

2. change the spin multiplicity.

Hope it helps!

1.) Using scf=xqc (Read: SCF - G09)

2.) Or using iop(5/13=1) (Read: G09 IOPs Manual)

use opt=(restart,maxcyc=200) , or start over with new input based on the non-optimized geometry, I mean the non optimized wavefunction will be the input of the new one

Hello Alžběta Türková,

I need your input because i am not getting that what you are actually trying to do....?

I guess you are doing oniom calculations as seen from you input.

use scf=qc

it is slow but gives results so it is better to optimize the maximum number of CPUs used %nprocshared=4 or 8,16 ...ect

scf=qc

scf=xqc

iop's etc are several ways to get rid of convergence related errors.

What is the unit of electrostatic potential (ESP) in the Gaussian09 output file, Hartree?

How to convert it to kJ/mole?

1 Hartree = 2625.5 kJ/mole ?

following

choose small best set to do the first optimisation and then re-optimised it

Applications of softwares is now on a higher trend. However, authors may highlight their work in any journal by -

1. Applying Graphics

2. Publish Open Access/Publish with good Journal Publishers

3. Excellent Review of respective Topic with systematic references

4. Have a strong network to share work on social platform

5. Present work in Conferences etc and get awarded

Can I help me ?I want to resolve this question of convergence in Gaussian calculation . I don't say what is the problem .

Initial guess = 0.0000 = 0.0000 = 0.5000 = 0.7696 S= 0.5097

Harris functional with IExCor= 402 diagonalized for initial guess.

ExpMin= 3.60D-02 ExpMax= 6.29D+03 ExpMxC= 9.49D+02 IAcc=3 IRadAn= 5 AccDes= 0.00D+00

HarFok: IExCor= 402 AccDes= 0.00D+00 IRadAn= 5 IDoV= 1

ScaDFX= 1.000000 1.000000 1.000000 1.000000

FoFCou: FMM=F IPFlag= 0 FMFlag= 100000 FMFlg1= 0

NFxFlg= 0 DoJE=T BraDBF=F KetDBF=T FulRan=T

Omega= 0.000000 0.000000 1.000000 0.000000 0.000000 ICntrl= 500 IOpCl= 0

NMat0= 1 NMatS0= 1 NMatT0= 0 NMatD0= 1 NMtDS0= 0 NMtDT0= 0

I1Cent= 4 NGrid= 0.

Petite list used in FoFCou.

Requested convergence on RMS density matrix=1.00D-08 within 999 cycles.

Requested convergence on MAX density matrix=1.00D-06.

Requested convergence on energy=1.00D-06.

No special actions if energy rises.

Alzbeta Tuerkova ,1st option you have to add cycle ,

second you have to use different scf mode for convergence and

third one is "you have to apply small move an atom on structure because sometimes atom will on trap ".

Hope it will help for you and those convergence failure victim.

If you are not using the ONIOM or QM/MM method for the protein system (this method is best to use Gaussview to set the high bottom layer and specify the appropriate atom type for the boundary H atom), it is a problem that the QM region optimization does not converge. Based on my years of experience using Gaussian, I recommend several commonly used methods

(1) For M06, M06-2X and other Minnesota series functionals, try to increase the accuracy of the functional integration grid points. G09 defaults to int=fine, which is equivalent to int=ultrafine (G16 defaults to this).

(2) For the case of using the dispersion function) It is recommended to try SCF=NoVarAcc first when there is no convergence.

(3) Use int=acc2e=12 to increase the integration accuracy.

(4) Use the SCF=conver=N keyword to change the convergence limit. G09/16 defaults to SCF=tight for single point calculation, which is equivalent to SCF=conver=8. However, the convergence requirements for the density matrix are a bit too strict at this time. It is often not easy to reach, and the energy is usually already converged to extremely high precision when it reaches it. For the purpose of calculating single point energy, you can rest assured to reduce to conver=6.

(5) Use quadratic convergence method (Quadratic convergence). This method requires fewer steps to converge than the default, and it can solve non-convergence to a greater extent. SCF=QC means using this method. In G09, SCF=XQC means to iterate to 64 rounds in a normal way (or set the value by scf=maxcyc), if it does not converge, then it will automatically switch to QC.

You only need to set up increasing the number of maximum cycle above 128 default cycles in the input file (SCF Done: E(RB3LYP) = -14532.8135584 A.U. after 129 cycles, it means that the result does not converge within 128 cycles). For example: Can you try ... opt=(Maxcyc=500) or SCF= (Maxcyc=500)....

if you remove integral keyword?

First, could you please try the keywords of opt=(Maxcyc=500) or SCF= (Maxcyc=500). Best regards.

Hello folks,

I'm getting the same error in QSite when I'm running the QM/MM. Anyone please help me to understand and solve this error?. The error is given below:

etot 99 N Y 1 U -1722.12820254192 8.9E-04 3.3E-05 3.0E-03

etot 100 N N 1 U -1722.12890662288 7.0E-04 0.0E+00 0.0E+00

ERROR 2256: fatal error

SCF convergence failure: Maximum number of iterations reached

Error: jaguar died in program scf

Your cooperation will highly be appreciated

Dear all RG colleagues,

Nice to share an outstanding summary about "How to solve the error message in Gaussian" as follow:

https://docs.computecanada.ca/wiki/Gaussian_error_messages

Have a good day!!!

Dang T. Nguyen this is good resource. Thank you.

I am running a job regarding CNT interaction with some compound by giving basic sets opt Freq wb97xd/6-31g(d) scf=xqc nonsym but the job is not running plse help me

You can also change the iop(2/9=2,5/6=6,5/7=999) say this was initial iop and convergence failure occurred. So, one can decrease the convergence criteria say (5/6=5) or (5/6=4).

You can use the SCF=XQC keyword. It works most of the time.

Put in the basis set, or remove the gen from the route line and specify an internal basis set.

optimize the structure at a lower level of theory and use that as the new starting structure...it may resolve the issue

In addition to all suggestions:

You can look at your unfinished job, ckeck and fix the structure, resolve the showed error, then, resumbit your job.