From owner-chemistry@ccl.net Tue Mar 24 23:34:01 2015 From: "Abrash, Samuel sabrash=-=richmond.edu" To: CCL Subject: CCL:G: Gaussian IR frequency calculation Message-Id: <-51195-150324225342-9921-x79Shsx3s/vXY0mgLfzaIQ-*-server.ccl.net> X-Original-From: "Abrash, Samuel" Content-Language: en-US Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="utf-8" Date: Wed, 25 Mar 2015 02:53:33 +0000 MIME-Version: 1.0 Sent to CCL by: "Abrash, Samuel" [sabrash,richmond.edu] All points well taken. Of course, the molecule must be one that is appropriately treated by single reference methods. In addition, as you point out, macromolecules, even small proteins, will be more difficult to treat. My source for the claim that anharmonicity is the bulk of the remaining error for those systems for which KS DFT is appropriate is a paper by Benny Gerber, in which he managed for sets of small and medium sized molecules to reduce the error in anharmonically corrected vibrational spectra to 1% or less. Again, Robert, your points are well-taken. ________________________________________ > From: owner-chemistry+sabrash==richmond.edu[-]ccl.net [owner-chemistry+sabrash==richmond.edu[-]ccl.net] on behalf of Robert Molt r.molt.chemical.physics#gmail.com [owner-chemistry[-]ccl.net] Sent: Tuesday, March 24, 2015 6:53 PM To: Abrash, Samuel Subject: CCL:G: Gaussian IR frequency calculation Sent to CCL by: Robert Molt [r.molt.chemical.physics*|*gmail.com] The issue is not whether it's 3%, the issue is the source, to me. You claim the error is anharmonicity. What evidence is there of that? For a molecule with very small modes (say, a protein), we are going to be down to 10 inverse centimeter modes. Those will be heavily anharmonically dominated. That can be a single-reference organic molecule, it's just big, and coupling of modes occurs. By contrast, a molecule which is significantly multi-reference will have its error having nothing to do with anharmonicity, but because multi-reference wavefunctions are often completely wrong when treated by KS-DFT (and many other methods). On 3/24/15 1:03 PM, Abrash, Samuel sabrash- -richmond.edu wrote: > Sent to CCL by: "Abrash, Samuel" [sabrash ~~ richmond.edu] > So I think the data that Robert just cited supports my claim of a 3% error. In the range 4000 cm-1 to 400 cm-1, the mid-ir, a 3% error would range from 120 cm-1 to 12 cm-1 so a 31 cm-1 average seems to be a bit below 3%. The other support for my claim is to go to the CCCDB (I think that's the correct number of C's) hosted at NIST, and check out the correction factors. For the types of high quality functional/basis set combinations, the correction factors are greater than .97, again consistent with an error of ~3%. > > Samuel A. Abrash > Associate Professor, Department of Chemistry > Coordinator, Environmental Studies Program > University of Richmond > sabrash~!~richmond.edu > Cell: (804) 363-2597 > www.richmond.edu/~sabrash > “At the time God created the world and desired to reveal the depth of His being from out of the hidden, the light came from darkness and they were joined together. Because of this, out of darkness came the light, and out of the hidden came the revealed and out of the good came evil and out of mercy came severe judgement, and everything is intertwined with everything else… the good inclination and the evil inclination, the right and the left.” The Zohar > > -----Original Message----- >> From: owner-chemistry+sabrash==richmond.edu~!~ccl.net [mailto:owner-chemistry+sabrash==richmond.edu~!~ccl.net] On Behalf Of Robert Molt r.molt.chemical.physics ~ gmail.com > Sent: Monday, March 23, 2015 10:58 PM > To: Abrash, Samuel > Subject: CCL:G: Gaussian IR frequency calculation > > > Sent to CCL by: Robert Molt [r.molt.chemical.physics]![gmail.com] > I respectfully disagree with the claim that KS-DFT functionals get 97% accuracy of the correct value, and the remaining 3% is mostly anharmonic. That assumes that KS-DFT functionals are describing electron correlation perfectly, as opposed to it being an approximate method varying in semi-empirical character from none to high. > > The mean absolute error over the G2 set of vibrational frequencies of B3LYP is 31 wavenumbers (see the 1998 paper from Adamo and Barone), obviously weighted more heavily for the higher values frequencies statistically. The RMS error can be as high as 42 wavenumbers (1993 Pople paper). > > The first harmonic mode of many molecules is often small, and the percent error is much higher on this first mode (independent of harmonic vs. anharmonic). > > On 3/23/15 8:51 PM, Abrash, Samuel sabrash%a%richmond.edu wrote: >> Sent to CCL by: "Abrash, Samuel" [sabrash|a|richmond.edu] With an >> appropriate functional with a good basis set, dft should get you to within about 97% of the correct value. The remainder of the error will be due to the harmonic approximation. Adding an anharmonic correction can get you to within 99%. >> >> Sent from my iPhone >> >>> On Mar 23, 2015, at 8:45 PM, zborowsk zborowsk^chemia.uj.edu.pl wrote: >>> >>> >>> Sent to CCL by: zborowsk [zborowsk^chemia.uj.edu.pl] W dniu >>> 2015-03-23 19:26, ggotelli()kaliumtech.com napisał(a): >>>> Sent to CCL by: [ggotelli[]kaliumtech.com] Thank you all for your >>>> help! >>>> I'll condense all my notes in the same mail. >>>> >>>> Krzysztof , yes I observed the difference in the whole spectrum. >>> Well, something strange. 300-400 cm-1 difference between theory and experiment around 3000 cm-1 and also around 100 cm-1?? By the way, do you have positive or negative (imaginary) frequencies around 100??? >>> >>> >>>> Serdar, the molecule that I 'm trying to calculate and where I have >>>> the differences is a dimer derived from dilactide. To estimate a >>>> scaling factor my reference molecule is dilactide in itself, which >>>> in turn has little differences between the calculated spectrum and >>>> the real one, depending on the peaks. >>>> >>>> André, thanks for the scaling factor reference. >>>> >>>> Stent, the polymer spectrum was made with an ATR device and without solvent. >>>> Because of this all calculations have done in vacuum. >>>> >>>> Robert, thanks for all your comments, they made me realize that my >>>> error was trying to adjust a dimer spectrum to a polymer spectrum, >>>> what are, of course, very different things. Although the dimer peaks >>>> are displaced with respect to those of the polymer, it can be seen >>>> from the GView animation that they correspond to what is expected, beyond the displacement involved. >>>> >>>> Regards! >>>> Gustavo Gotelli >>>> Polymer Synthesis >>>> Pharmaceutical Technology Department Faculty of Pharmacy and >>>> Biochemistry Buenos Aires University Argentina >>>> >>>> -----Mensaje original----- >>>> De: owner-chemistry+ggotelli==kaliumtech.com a ccl.net >>>> [mailto:owner-chemistry+ggotelli==kaliumtech.com a ccl.net] En >>>> nombre de Víctor Luaña Cabal victor^^fluor.quimica.uniovi.es Enviado >>>> el: sábado, 21 de marzo de 2015 21:16 >>>> Para: Gotelli, Gustavo A >>>> Asunto: CCL: Gaussian IR frequency calculation >>>> >>>> >>>> Sent to CCL by: =?iso-8859-1?Q?V=EDctor_Lua=F1a?= Cabal >>>> [victor]|[fluor.quimica.uniovi.es] >>>> On Sat, Mar 21, 2015 at 05:56:44PM -0400, Robert Molt >>>> r.molt.chemical.physics],[gmail.com wrote: >>>>> Sent to CCL by: Robert Molt [r.molt.chemical.physics---gmail.com] >>>>> Moreover, you are using the B3LYP functional. It has limited >>>>> accuracy, depending on the system specifics. If it is a >>>>> multi-reference problem, great care is necessary. >>>>> >>>>> Also note that you can only compare the TOTAL spectrum, not any one peak. >>>> Robert, >>>> >>>> 1) Let me take your highly correct advise to remember to the newest >>>> generations that the role on computational chemistry is not being >>>> able to reproduce the number of the experiment. >>>> >>>> 2) The experiment helps in modulating our calculations, and the final >>>> objective is a perfect reprodution of the experiment. >>>> >>>> 3) But the reason is not substituting experiments. The reason is that >>>> in the processs of mimicking it we learn the rules that govern >>>> Nature. >>>> >>>> 4) There are too many questions in this forum that tend to ignore >>>> this rule and it is important to remember it again and again. >>>> >>>> Best regards, >>>> Dr. Víctor Luaña >>>> -- >>>> . . "The hardest part in solving a problem is recognizing >>>> / `' \ its existence. Learning the causes CAN be the road to >>>> /(o)(o)\ the solution." >>>> /`. \/ .'\ -- ¿? >>>> / '`'` \ "Lo mediocre es peor que lo bueno, pero también es peor >>>> | \'`'`/ | que lo malo, porque la mediocridad no es un grado, es >>>> |una | '`'`| | actitud" >>>> \/`'`'`'\/ -- Jorge Wasenberg, 2015 >>>> ===(((==)))==================================+========================= >>>> ! Dr.Víctor Luaña ! Mediocre is worse than >>>> ! Departamento de Química Física y Analítica ! good, but it is also ! >>>> Universidad de Oviedo, 33006-Oviedo, Spain ! worse than bad, because >>>> ! e-mail: victor-#-fluor.quimica.uniovi.es ! mediocrity is not a grade, >>>> ! phone: +34-985-103491 fax: +34-985-103125 ! it is an attitude >>>> +--------------------------------------------+ >>>> GroupPage : http://azufre.quimica.uniovi.es/ (being >>>> reworked)http://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.c >>>> cl.net/chemistry/sub_unsub.shtmlhttp://www.ccl.net/spammers.txt-- >>> https://www.youtube.com/watch?v=IGyfx3Pmeyk >>> Krzysztof K. Zborowski >>> Faculty of Chemistry >>> Jagiellonian University >>> 3 Ingardena Street >>> 30-060 Krakow >>> Poland >>> phone: +48(12)663-2067 >>> email: zborowsk(~)chemia.uj.edu.pl >>> gg 3817259 >>> skype kzys70 >>> www.chemia.uj.edu.pl/~zborowskhttp://www.ccl.net/chemistry/sub_unsub. >>> shtml> > -- > Dr. Robert Molt Jr. > r.molt.chemical.physics=-=gmail.com > Nigel Richards Research Group > Department of Chemistry & Chemical Biology Indiana University-Purdue University Indianapolis LD 326 > 402 N. Blackford St. > Indianapolis, IN 46202http://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.ccl.net/chemistry/sub_unsub.shtmlhttp://www.ccl.net/spammers.txt> > -- Dr. Robert Molt Jr. r.molt.chemical.physics]^[gmail.com Nigel Richards Research Group Department of Chemistry & Chemical Biology Indiana University-Purdue University Indianapolis LD 326 402 N. Blackford St. Indianapolis, IN 46202http://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.ccl.net/chemistry/sub_unsub.shtmlhttp://www.ccl.net/spammers.txt