From owner-chemistry@ccl.net Sun Nov 27 19:13:00 2011 From: "Saurabh S Chitnis sschitnis%x%gmail.com" To: CCL Subject: CCL: questions on convergence, restarting jobs, counterpoise Message-Id: <-45929-111127170210-2902-aTpKAAVAUeNvfKwPyi46Iw-$-server.ccl.net> X-Original-From: "Saurabh S Chitnis" Date: Sun, 27 Nov 2011 17:02:07 -0500 Sent to CCL by: "Saurabh S Chitnis" [sschitnis _ gmail.com] Hello, I have three questions: 1. In many of my "opt" and "freq=raman" jobs, I find that the calculation completes successfully with the forces having converged but the displacements still being above the convergence threshold. How can this be fixed? I usually face this problem when using opt=verytight and int=ultrafine. 2. Is there any way of restarting an analytic frequency job which has completed but not converged? 3. Can the output of a counterpoise calculation (i.e. the Corrected total energy) be interpreted as a bond-dissociation energy between the two fragments? Here are some details regarding the above questions: theory: DFT (B3LYP, PBE1PBE and B3PW91) basis-set: cc-pvtz and aug-cc-pvtz systems being calcualated: P2Me4, [P2Me5]+ and [P2Me6]2+ for the couterpoise: I'm trying to get the P-P bond dissocation energy. So I optimize the geometry as per usual and then redefine the optimized P2Me4 as fragment1(Me2P) and fragment2(PMe2) to run the counterpoise calculation. The result should be the homolytic bond dissociation energy right? cheers Saurabh From owner-chemistry@ccl.net Sun Nov 27 19:48:01 2011 From: "Josh Marell mare0051(a)umn.edu" To: CCL Subject: CCL: Convergence During Frequency Analysis Message-Id: <-45930-111127192934-28069-66xJgG2CraHJ45FhYyoAqg!=!server.ccl.net> X-Original-From: Josh Marell Content-Type: multipart/alternative; boundary=20cf300e568b4889f804b2c098bc Date: Sun, 27 Nov 2011 18:29:04 -0600 MIME-Version: 1.0 Sent to CCL by: Josh Marell [mare0051 * umn.edu] --20cf300e568b4889f804b2c098bc Content-Type: text/plain; charset=ISO-8859-1 Hi Cina, Thank you for the response. I rechecked the calculations, and have switched to instead doing an initial optimization with opt=caclfc (to get closer to the real minimum), and then a second step of opt=calcall, which usually finishes in just one or two extra optimization steps along with the frequency analysis output. Thanks for the assistance, Josh On Sun, Nov 13, 2011 at 11:52 PM, cina foroutan canyslopus|yahoo.co.uk < owner-chemistry,,ccl.net> wrote: > Dear Josh, > > I have had this problem many times before. This tells that your geometry > does not correspond to the real local minimum. To check it you can submit a > new optimization with your optimized geometry and consider opt=readfc > guess=read keywords (please note that if you saved a chk file you can use > these keywords!). If your structure is the real minimum, the new job must > finishes after just one cycle. However, I guess that, this is not the case > for you. > To get the real minimum on the PES you may also consider opt=(calcall). > Then, the second derivatives of energy will be computed in every step of > optimization and you can be sure that you will find the real local minimum. > Unfortunately, calcall keyword is very demanding because as I mentioned in > every step of the optimization second derivatives of energy must be > computed, this means that every optimization step is as demanding as a > frequency calculation. > > Good luck > Cina Foroutan-Nejad > > ------------------------------ > *From:* Josh Marell mare0051-#-umn.edu > *To:* "Foroutan-Nejad, Cina " > *Sent:* Monday, 14 November 2011, 1:33 > *Subject:* CCL: Convergence During Frequency Analysis > > Hello, > > I am optimizing a geometry at the m062x/6-31+g(d,p) level, and then I > perform a frequency analysis. During the optimization step (and utilizing > opt=verytight), I ensure that I am getting full convergence and see: > > Maximum Force 0.000000 0.000002 YES > RMS Force 0.000000 0.000001 YES > Maximum Displacement 0.000005 0.000006 YES > RMS Displacement 0.000001 0.000004 YES > > However, then I carry out a frequency analysis using the following route > section: > # m062x/6-31+g(d,p) freq integral(ultrafinegrid) scrf=check geom=checkpoint > > And at the end of the output file for the frequency analysis, it indicates > in this specific case that none of the criteria have converged (this is > after the line that specifies the axes has been restored to the original > set) > > Item Value Threshold Converged? > Maximum Force 0.024599 0.000450 NO > RMS Force 0.005725 0.000300 NO > Maximum Displacement 0.311758 0.001800 NO > RMS Displacement 0.083739 0.001200 NO > Predicted change in Energy=-4.741844D-03 > GradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGrad > > Is this telling me the results of the frequency analysis are not reliable? > I guess I'm unclear why the geometry that converged with verytight > requirement is now not converging during the frequency job (assuming I'm > understanding the implication of the above correctly). > > Thank you, > > Josh > > > --20cf300e568b4889f804b2c098bc Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Hi Cina,

Thank you for the response. =A0I rechecked the = calculations, and have switched to instead doing an initial optimization wi= th opt=3Dcaclfc (to get closer to the real minimum), and then a second step= of opt=3Dcalcall, which usually finishes in just one or two extra optimiza= tion steps along with the frequency analysis output.

Thanks for the assistance,

Jos= h

On Sun, Nov 13, 2011 at 11:52 PM, cina foroutan = canyslopus|yahoo.co.uk <owner-chemistry,,ccl.net= > wrote:
Dear Josh,

I = have had this problem many times before. This tells that your geometry does= not correspond to the real local minimum. To check it you can submit a new= optimization with your optimized geometry and consider opt=3Dreadfc guess= =3Dread keywords (please note that if you saved a chk file you can use thes= e keywords!). If your structure is the real minimum, the new job must finis= hes after just one cycle. However, I guess that, this is not the case for y= ou.
To get the real minimum on the PES you may also consider opt=3D(= calcall). Then, the second derivatives of energy will be computed in every = step of optimization and you can be sure that you will find the real local = minimum. Unfortunately, calcall keyword is very demanding because as I ment= ioned in every step of the optimization second derivatives of energy must be com= puted, this means that every optimization step is as demanding as a frequen= cy calculation.

Good luc= k
Cina Foroutan-Nejad

From: Josh Marell mare0051-#-umn.edu <owner-chemistry{}ccl.net>
To: "Foroutan-Nejad, Ci= na " <canyslopus{}= yahoo.co.uk>
Sent:= Monday, 14 November 2011, 1:33
Subject: CCL: Convergence Du= ring Frequency Analysis

Hello,

I = am optimizing a geometry at the m062x/6-31+g(d,p) level, and then I perform= a frequency analysis. =A0During the optimization step (and utilizing opt= =3Dverytight), I ensure that I am getting full convergence and see:

Maximum Force =A0 =A0 =A0 =A0 =A0 =A00.000000 =A0 = =A0 0.000002 =A0 =A0 YES
=A0RMS =A0 =A0 Force =A0 =A0 =A0 =A0 =A0= =A00.000000 =A0 =A0 0.000001 =A0 =A0 YES
=A0Maximum Displacement= =A0 =A0 0.000005 =A0 =A0 0.000006 =A0 =A0 YES
=A0RMS =A0 =A0 Displacement =A0 =A0 0.000001 =A0 =A0 0.000004 =A0 =A0 YES

However, then I carry out a frequency analysi= s using the following route section:
# m062x/6-31+g(d,p) freq int= egral(ultrafinegrid) scrf=3Dcheck geom=3Dcheckpoint

And at the end of the output file for the frequency ana= lysis, it indicates in this specific case that none of the criteria have co= nverged (this is after the line that specifies the axes has been restored t= o the original set)

=A0 =A0 =A0 =A0 =A0Item =A0 =A0 =A0 =A0 =A0 =A0 = =A0 Value =A0 =A0 Threshold =A0Converged?
=A0Maximum Force =A0 =A0 =A0 =A0 =A0 =A00.024599 =A0 =A0 0.000450 =A0 = =A0 NO=A0
=A0RMS =A0 =A0 Force =A0 =A0 =A0 =A0 =A0 =A00.005725 = =A0 =A0 0.000300 =A0 =A0 NO=A0
=A0Maximum Displacement =A0 =A0 0.= 311758 =A0 =A0 0.001800 =A0 =A0 NO=A0
=A0RMS =A0 =A0 Displacement= =A0 =A0 0.083739 =A0 =A0 0.001200 =A0 =A0 NO=A0
=A0Predicted change in Energy=3D-4.741844D-03
=A0GradGradGra= dGradGradGradGradGradGradGradGradGradGradGradGradGradGradGrad

Is this telling me the results of the frequency ana= lysis are not reliable? =A0I guess I'm unclear why the geometry that co= nverged with verytight requirement is now not converging during the frequen= cy job (assuming I'm understanding the implication of the above correct= ly).

Thank you,

Josh



--20cf300e568b4889f804b2c098bc-- From owner-chemistry@ccl.net Sun Nov 27 21:13:01 2011 From: "Theodore S. Dibble tsdibble*o*esf.edu" To: CCL Subject: CCL: Suitable Basis set for Hg(II) Calculation Message-Id: <-45931-111127211154-26043-C9CnzfAE9dRADnwXSDJZtA-,-server.ccl.net> X-Original-From: "Theodore S. Dibble" Date: Sun, 27 Nov 2011 21:11:51 -0500 Sent to CCL by: "Theodore S. Dibble" [tsdibble##esf.edu] Pezhman, In general, one expects the LANL2DZ basis set is far surpassed by the Stuttgart basis sets, available at http://www.theochem.uni-stuttgart.de/pseudopotentials/index.en.html. This site provides ECPs and matching correlation-consistent basis sets for the outer electrons. It is certainly possible that LANL2DZ actually performs better than more recent basis sets for certain classes of problems. Are there experimental data sets you can use to validate your methods? Best wishes, Ted Dibble Theodore S. Dibble Professor of Chemistry SUNY-Environmental Science and Forestry 1 Forestry Drive Syracuse, NY 13210 (315) 470-6596 (315) 470-6856 (fax) http://www.esf.edu/chemistry/faculty/dibble.htm > "Pezhman Zarabadi-Poor pzarabadip]~[gmail.com" wrote: > > Sent to CCL by: "Pezhman Zarabadi-Poor" [pzarabadip/a\gmail.com] > Dear CCL users, > > I am planning to do some DFT and TDDFT calculations on a complex of Hg(II) and > a naphthalene sulfonic acid derivative. I searched through the literature and > found that some of previous computational works have employed ECP-LANL2DZ for > Hg(II). Is it a good suggestion for doing such calculations? I checked the > larger basis sets such as aug-cc-pVDZ to QZ but I obtained a little bit > different results than the published ones? According to the consultant with an > expert person, there is guess that maybe employing larger basis sets makes Hg > ions too soft and makes the results wrong! > > Your suggestions about the appropriate method/basis sets are highly > appreciated. > > Best regards, > Pezhman Zarabadi-Poor > >