Recent Posts

Pages: [1] 2 3 ... 10
1
Hi Martijn,

hm, this seems to be unrelated to the G0W0 step, rather it may be an error while reading in/processing the auxiliary basis set. Can you check the beginning of the output for the basis section; specifically the last occurrence of this one:

 AUXILIARY BASIS SET information:

              we will work with the 1s 3p 5d 7f 9g ... basis set
              ...i.e. with spherical basis functions...

   type   atoms  prim   cont   basis
   ---------------------------------------------------------------------------
    cd       1    131    131   x2c-SVPall-2c   [25s5p5d3f5g|25s5p5d3f5g]
   ---------------------------------------------------------------------------
   total:    1    131    131
   ---------------------------------------------------------------------------

and check if the readout there makes sense in your calculation? Unfortunately I could not reproduce this error with any version I have available, and this error is not familiar to me. But the last line printed out (lit,ljt,lkt: ) suggests that the auxiliary basis set is weird/invalid - the numbers report f-orbitals for the orbital basis set ( first two numbers = 4, matching for the x2c-SVPall-2c basis set) but 0 for the auxiliary index, which means lower than an s-orbital. Maybe something went wrong using the stream editor :/.

All the best,
Christof
2
Thanks Christof. That works other than that next the G0W0 calculation crashes with the following error message:

           ----------------------
               RI - INFORMATION
           ----------------------

 biggest AO integral is expected to be    32.038321512

 Threshold for integral neglect (rithr2):   0.89E-11
 Threshold for integral neglect (rithr1):   0.89E-11

 Contributions to RI integral batches:
  neglected integral batches:                  1364
  direct contribution:                            0
  memory contribution:                  4964
 Core memory available (ricore)            500 MiB
 Core memory needed for (P|Q) and Cholesky 0 MiB
 Core memory used for integrals            0 MiB
 
 ****************************************
 Memory allocated for RIDFT 0 MiB
 ****************************************
                                           
  lit,ljt,lkt:                     4                     4                     0

========================
 internal module stack:
------------------------
    escf
    lpdrec
========================

 l-qun in <asrstat> messy
 escf ended abnormally


Do you know what the origin of that might be? I assume that is not because of using the jbas for the cbas.

Thanks,

Martijn

P.S. this is the GW bit of control:

$scfinstab soghf
$soes all
$rpacor   22000
$rick
$rigw
  rpa
  mxdiis       8
  npoints    128
  npade      128
  rpoints     16
  rshift    0.20000000
  ips+     1
  gap
  eta       0.00100000
3
Hi Martijn,

for the all-electron x2c-SVPall-2c basis sets specialized cbas-type auxiliary basis sets unfortunately do not yet exist (as far as I know however they are being worked on, but not by me. : ) ). The quickest solution is therefore to use the jbas also as cbas. This has to be done manually as of version 7.5. Under linux these two command lines should achieve this:

sed -ie 's/jbas/cbas/g' auxbasis
sed -ie 's/jbas/cbas/g' control

They change all occurrences of "jbas" in the auxbasis and control file to "cbas". For escf alone this is enough. In case you want to re-add the according jbas (to be able to run ridft in case it was not already done), just proceed a second time through define and the "ri" menu in the final section. This will re-add the jbas. As the jbas basis for the x2c basis set is fairly large I do not expect enlarged errors by using the jbas also in the GW procedure.

In case you perform two-component calculations using the dhf-SVP-2c (or higher) basis set family, using the def2-SVP/def2-TZVP etc cbas is fine.

All the best,
Christof
4
Escf and Egrad / cbas for relativistic two-component all-electron GW calculation
« Last post by martijn on May 07, 2021, 12:17:27 PM »
Hi,

If I want to run a relativistic two-component all-electron GW calculation on a molecule containing elements for which the normal basis-set has an ECP, e.g. Cd, how do I choose the cbas?

For the ground-state DFT (using $soghf, $kramers and $rx2c) I use x2c-SVPall-2c as basis and that basis-set is also available for jbas but not cbas.

Thanks,

Martijn
5
Miscellaneous / Alternate SCF algorithms?
« Last post by jad1507 on May 07, 2021, 12:05:20 PM »
Hello Turbomole Hive Mind,

I am having problems converging Aufbau (within the beta manifold) electronic configurations for some high spin complexes I am looking at. Manually changing the electronic structure to be Aufbau and then reoptimising merely results in a non-Aufbau system again. I know that on other quantum chemical packages there are other SCF algorithms that can be used that might resolve this issue I'm having: so I am wondering if Turbomole has anything similar? I've consulted the manual but not had much luck.

Thanks in advance,
jad1507
6
TmoleX - general topics / Re: configuring TmoleX for queue that runs on local system
« Last post by KrisR on April 23, 2021, 11:07:41 AM »
About fooling...
I found this directory `/opt/TmoleX2021/extapps/putty`and there were two programs `pscp`and `pssh`.
I tried to use `pssh` to connect to another server. It writes what user-name will be used but no connection is built.
Also `pscp` does not work -- the only message when I tried to copy a file is "Lost connection".

I replaced both these binaries with links to /usr/bin/ssh and /usr/bin/scp and then I've seen that the job was submitted to queue and eventually calculation has succeeded.
I have used the same directory and it seems that TmoleX has no problem with that. It just created additional temporary directory that was eventually removed.
So it seems at first glance that something is wrong with compiled versions of plink and pscp on linux.

I'm happy - it works :-D
7
Escf and Egrad / Re: qsGW vs evGW convergence
« Last post by martijn on April 23, 2021, 10:45:55 AM »
Thanks Christof
8
Escf and Egrad / Re: qsGW vs evGW convergence
« Last post by chris.hol on April 22, 2021, 11:55:07 PM »
Hi,

indeed this is to be expected; qsGW will in general need many more iterations than evGW and sometimes even more than HF/DFT calculations. qsGW updates all the Fock-Matrix elements, making a re-diagonalization of the Fock-Matrix necessary in every iteration. Especially in cases were the self-energy has peaks close to the Fock-Matrix elements a steep dependence of the latter to small changes can be expected. This leads to a rather slow convergence, and DIIS is not helping too much in these cases. In evGW only the eigenvalues/quasiparticle energies but not the eigenvectors of the Fock-Matrix are updated, a re-diagonalization is not needed. Therefore, it converges pretty fast especially with DIIS.

All the best,
CHristof

9
TmoleX - general topics / Re: configuring TmoleX for queue that runs on local system
« Last post by KrisR on April 22, 2021, 06:22:32 PM »
Hi Uwe,
I'll try to answer questions in order.

1st of all: "why-question".
Well, we are here mostly preparative institute and ph.d.-students using QM to compare or to confirm some spectroscopic data. We have a few post-docs that are on higher level but majority of users misuse QM ;-)
Most students use gaussian/gaussview which they have installed both in cluster and on their PCs. They are used to open their results "as it goes". If they are in some windows directory they start windows versions if they are on cluster and they were checking log-files they starting gaussview on cluster.
I really would love to make them more aware of capabilities of turbomole and tmolex plays there huge role. But they are chemists working in lab not in silico. They need to see results.
For that reason I've not only installed turbomole on our cluster but also tmolex.
In hands-on seminars I'm explaining how to configure TmoleX for remote queue but there is always danger that they will start it on cluster. If a few of them will start local computations on head-node we will be not able to log-in.
For that reason I'd like to explain them how to configure TmoleX to work on cluster.

I tried the option with different directory and it did not worked. I didn't have much hope as the test connection did not worked but I tried. I couldn't see anything it TmoleX and also qstat didn't show a submitted job.

With fooling TmoleX -- I'll try it tomorrow and I will let you know.

Thanks for response, Kris
10
TmoleX - general topics / Re: configuring TmoleX for queue that runs on local system
« Last post by uwe on April 22, 2021, 03:25:30 PM »
Hi,

it is quite unusual that the server of a cluster is at the same time a desktop machine. If you usually submit the job from your Windows machine to the queueing system, and as I guess that your Windows machine is not the head node of the cluster, why do you want to submit jobs from a remote machine to a remote machine in addition or as an alternative? OK, sorry, questions about 'why' are more out of curiosity and not helpful as answers...

If ssh to the same machine works, the Run(network) option should work too. What comes to my mind when thinking about possible problems is the fact that the local directory and the remote directory must not be the same. Please try to set as working directory on the remote machine a different path than the one of your project.

Did you enable 'Expert  settings' in the remote systems dialog and set a check at 'no password'?

Will you ALWAYS be submitting jobs to the local queuing system and not use another remote machine or cluster from this TmoleX client on that machine? If yes, you could try to fool TmoleX. Replace the ssh and scp binaries that you will find in the local TmoleX installation by scripts which simply do nothing but execute the commands that are given as options. It will be a bit of a puzzle to find out how to do that exactly (e.g. Tmolex will be called 'ssh user@machine mkdir /home/user/TmoleXProject/job_SP_1', so your ssh script has to ignore the first option and execute the rest locally. Same for scp which has to be replaced by a simple 'cp' but ignores the machine and the user name.).

Regards,
Uwe
Pages: [1] 2 3 ... 10