From owner-chemistry@ccl.net Tue Nov 16 07:22:00 2021 From: "Tobias Kraemer tobias.kraemer.:.mu.ie" To: CCL Subject: CCL:G: SCF convergence in G16 Message-Id: <-54496-211116071954-4049-hzCUM/j4bSjDwyhwPGmofA-#-server.ccl.net> X-Original-From: "Tobias Kraemer" Date: Tue, 16 Nov 2021 07:19:49 -0500 Sent to CCL by: "Tobias Kraemer" [tobias.kraemer _ mu.ie] Hello everyone, I have a question regarding SCF convergence in G16 (I am using revision B.01, but probably not relevant). I've noticed that during a scan the calculation terminates with an error due to SCF non-convergence. I use a smaller SCF convergence to speed up the calculation, and here use scf=(conver=4) However, this setting seems to be overruled by the SCFChk, which reports SCFChk: SCF convergence is xxxx required 1.0D10-08 I have not seen this before with Gaussian 09 and did not find much information about this online (some report same issue). Has anyone any advice what to do in order to de-activate this SCFChk? I am surely not the first person to experience this problem. Thanks in advance. Best wishes Tobias From owner-chemistry@ccl.net Tue Nov 16 08:28:00 2021 From: "Frank Jensen frj###chem.au.dk" To: CCL Subject: CCL:G: SCF convergence in G16 Message-Id: <-54497-211116082500-29327-8umdSKRZvKtSROALMOt6Ww%a%server.ccl.net> X-Original-From: Frank Jensen Content-Language: en-US Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="us-ascii" Date: Tue, 16 Nov 2021 13:24:31 +0000 MIME-Version: 1.0 Sent to CCL by: Frank Jensen [frj . chem.au.dk] G16 has tightened the SCF convergence check compared to G09. SCFChk is called in link 701 and 801, which calculates integral derivatives and does the AO to MO transformation, respectively. The SCChk can be turned off by the undocumented iop 127 and 117, setting these to a value larger than -100 i.e. if l701 is causing the problem (likely the case for you) add the keyword iop(7/127=-99), if l801 is the problem, add keyword iop(8/117=-99) Frank -----Original Message----- > From: owner-chemistry+frj==chem.au.dk : ccl.net On Behalf Of Tobias Kraemer tobias.kraemer.:.mu.ie Sent: 16. november 2021 13:20 To: Frank Jensen Subject: CCL:G: SCF convergence in G16 Sent to CCL by: "Tobias Kraemer" [tobias.kraemer _ mu.ie] Hello everyone, I have a question regarding SCF convergence in G16 (I am using revision B.01, but probably not relevant). I've noticed that during a scan the calculation terminates with an error due to SCF non-convergence. I use a smaller SCF convergence to speed up the calculation, and here use scf=(conver=4) However, this setting seems to be overruled by the SCFChk, which reports SCFChk: SCF convergence is xxxx required 1.0D10-08 I have not seen this before with Gaussian 09 and did not find much information about this online (some report same issue). Has anyone any advice what to do in order to de-activate this SCFChk? I am surely not the first person to experience this problem. Thanks in advance. Best wishes Tobiashttp://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.ccl.net/chemistry/sub_unsub.shtmlhttp://www.ccl.net/spammers.txt From owner-chemistry@ccl.net Tue Nov 16 11:38:00 2021 From: "Tobias Kraemer tobias.kraemer::mu.ie" To: CCL Subject: CCL: SCF convergence issue G16 Message-Id: <-54498-211116113724-21879-opXYHdAC9/cnXQr+42wzcA===server.ccl.net> X-Original-From: "Tobias Kraemer" Date: Tue, 16 Nov 2021 11:37:20 -0500 Sent to CCL by: "Tobias Kraemer" [tobias.kraemer^^^mu.ie] Dear Frank, Thanks for your quick response this is most helpful. I had no clue about this new convergence check. Thanks for sharing these IOps, I will try it out. Best wishes, Tobias From owner-chemistry@ccl.net Tue Nov 16 12:13:01 2021 From: "Igors Mihailovs igorsm(!)cfi.lu.lv" To: CCL Subject: CCL:G: SCF convergence in G16 Message-Id: <-54499-211116111200-20059-ccM7qnPUsdapsqjNuymIwQ|*|server.ccl.net> X-Original-From: Igors Mihailovs Content-Language: lv Content-Type: multipart/alternative; boundary="------------848AEDEADB1BB20ACB6A5ECD" Date: Tue, 16 Nov 2021 18:11:44 +0200 MIME-Version: 1.0 Sent to CCL by: Igors Mihailovs [igorsm##cfi.lu.lv] This is a multi-part message in MIME format. --------------848AEDEADB1BB20ACB6A5ECD Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Dr. Jensen has already given the actual answer but I wanted to comment that after some tests I had to conclude that Gaussian 16 indeed does not even "listen" to SCF(Conver=N) _until You add also SCF=Sleazy _. I tested plain SCF for a water molecule, no scan. No other IOps documented in the official guide can make the convergence to untighten: I tried in a monkey fashion 5/6=4, 5/32=3, 5/82=4, 5/87=4; the only way to get the convergence up to D-04 was to add SCF=(Conver=2,Sleazy). I don't know why You should keep it twice the actual value. Maybe this is just the coincidence, maybe has something to do with what is said on Gaussian's home page: Conver=N Sets the SCF convergence criterion to 10^-N . SCF convergence requires both <10^-N RMS change in the density matrix and <10^-(N-2) maximum change in the density matrix. Note that the energy change is not used to test convergence; however, an SCF 10^-N RMS density matrix change typically corresponds to a 10^-2N change in energy in atomic units. [..] Note the N-2. *_________* Yours sincerely, Igors Mihailovs Research assistant Laboratory of Organic Materials Institute of Solid State Physics University of Latvia On 16.11.21 15:24, Frank Jensen frj###chem.au.dk wrote: > Sent to CCL by: Frank Jensen [frj . chem.au.dk] > G16 has tightened the SCF convergence check compared to G09. > SCFChk is called in link 701 and 801, which calculates integral derivatives and does the AO to MO transformation, respectively. > The SCChk can be turned off by the undocumented iop 127 and 117, setting these to a value larger than -100 > i.e. if l701 is causing the problem (likely the case for you) add the keyword iop(7/127=-99), if l801 is the problem, add keyword iop(8/117=-99) > > Frank > > -----Original Message----- >> From: owner-chemistry+frj==chem.au.dk() ccl.net On Behalf Of Tobias Kraemer tobias.kraemer.:.mu.ie > Sent: 16. november 2021 13:20 > To: Frank Jensen > Subject: CCL:G: SCF convergence in G16 > > > Sent to CCL by: "Tobias Kraemer" [tobias.kraemer _ mu.ie] Hello everyone, > > > I have a question regarding SCF convergence in G16 (I am using revision B.01, but probably not relevant). I've noticed that during a scan the calculation terminates with an error due to SCF non-convergence. I use a smaller SCF convergence to speed up the calculation, and here use scf=(conver=4) However, this setting seems to be overruled by the SCFChk, which reports > > SCFChk: SCF convergence is xxxx required 1.0D10-08 > > I have not seen this before with Gaussian 09 and did not find much information about this online (some report same issue). Has anyone any advice what to do in order to de-activate this SCFChk? I am surely not the first person to experience this problem. > > Thanks in advance. > > Best wishes > > Tobiashttp://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.ccl.net/chemistry/sub_unsub.shtmlhttp://www.ccl.net/spammers.txt> > E-mail to subscribers:CHEMISTRY.:.ccl.net or use:> > E-mail to administrators:CHEMISTRY-REQUEST.:.ccl.net or use> > Before posting, check wait time at:http://www.ccl.net > > Job:http://www.ccl.net/jobs > Conferences:http://server.ccl.net/chemistry/announcements/conferences/ > > Search Messages:http://www.ccl.net/chemistry/searchccl/index.shtml> > RTFI:http://www.ccl.net/chemistry/aboutccl/instructions/ > > --------------848AEDEADB1BB20ACB6A5ECD Content-Type: text/html; charset=utf-8 Content-Transfer-Encoding: 7bit Dr. Jensen has already given the actual answer but I wanted to comment that after some tests I had to conclude that Gaussian 16 indeed does not even "listen" to SCF(Conver=N) _until You add also SCF=Sleazy _. I tested plain SCF for a water molecule, no scan. No other IOps documented in the official guide can make the convergence to untighten: I tried in a monkey fashion 5/6=4, 5/32=3, 5/82=4, 5/87=4; the only way to get the convergence up to D-04 was to add SCF=(Conver=2,Sleazy). I don't know why You should keep it twice the actual value. Maybe this is just the coincidence, maybe has something to do with what is said on Gaussian's home page:

Conver=N

Sets the SCF convergence criterion to 10-N. SCF convergence requires both <10-N RMS change in the density matrix and <10-(N-2) maximum change in the density matrix. Note that the energy change is not used to test convergence; however, an SCF 10-N RMS density matrix change typically corresponds to a 10-2N change in energy in atomic units. [..]

Note the N-2.

_________
Yours sincerely,
Igors Mihailovs
Research assistant
Laboratory of Organic Materials
Institute of Solid State Physics
University of Latvia

On 16.11.21 15:24, Frank Jensen frj###chem.au.dk wrote:
Sent to CCL by: Frank Jensen [frj . chem.au.dk]
G16 has tightened the SCF convergence check compared to G09.
SCFChk is called in link 701 and 801, which calculates integral derivatives and does the AO to MO transformation, respectively.
The SCChk can be turned off by the undocumented iop 127 and 117, setting these to a value larger than -100
i.e. if l701 is causing the problem (likely the case for you) add the keyword iop(7/127=-99), if l801 is the problem, add keyword iop(8/117=-99)

Frank

-----Original Message-----

From: owner-chemistry+frj==chem.au.dk() ccl.net <owner-chemistry+frj==chem.au.dk() ccl.net> On Behalf Of Tobias Kraemer tobias.kraemer.:.mu.ie

Sent: 16. november 2021 13:20
To: Frank Jensen <frj() chem.au.dk>
Subject: CCL:G: SCF convergence in G16


Sent to CCL by: "Tobias  Kraemer" [tobias.kraemer _ mu.ie] Hello everyone,


I have a question regarding SCF convergence in G16 (I am using revision B.01, but probably not relevant). I've noticed that during a scan the calculation terminates with an error due to SCF non-convergence. I use a smaller SCF convergence to speed up the calculation, and here use scf=(conver=4) However, this setting seems to be overruled by the SCFChk, which reports 

SCFChk: SCF convergence is xxxx  required 1.0D10-08

I have not seen this before with Gaussian 09 and did not find much information about this online (some report same issue). Has anyone any advice what to do in order to de-activate this SCFChk? I am surely not the first person to experience this problem.

Thanks in advance.

Best wishes

Tobiashttp://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.ccl.net/chemistry/sub_unsub.shtmlhttp://www.ccl.net/spammers.txtE-mail to subscribers: CHEMISTRY.:.ccl.net or use:
      http://www.ccl.net/cgi-bin/ccl/send_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST.:.ccl.net or use
      http://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.ccl.net/chemistry/sub_unsub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.net/jobs 
Conferences: http://server.ccl.net/chemistry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/searchccl/index.shtmlhttp://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/instructions/

--------------848AEDEADB1BB20ACB6A5ECD--