From chemistry-request(+ at +)ccl.net Fri Aug 8 12:57:09 2003 Received: from yakko.cimav.edu.mx (yakko.cimav.ecustomer.mx [148.223.46.2] (may be forged)) by server.ccl.net (8.12.8/8.12.8) with ESMTP id h78Gv898012262 for ; Fri, 8 Aug 2003 12:57:08 -0400 Received: from yakko (localhost [127.0.0.1]) by localhost.cimav.edu.mx (Postfix) with ESMTP id D8B637611B; Fri, 8 Aug 2003 10:57:10 -0600 (MDT) Received: from localhost ([127.0.0.1]) by yakko.cimav.edu.mx (MailMonitor for SMTP v1.2.2 ) ; Fri, 8 Aug 2003 10:57:10 -0600 (MDT) Received: from laquicom02 (byron.cimav.edu.mx [148.223.46.1]) by yakko.cimav.edu.mx (Postfix) with SMTP id 4EA3576116; Fri, 8 Aug 2003 10:57:10 -0600 (MDT) Message-ID: <00b801c35dc5$b5392ef0$8400000a $#at#$ laquicom02> From: "Dr. Daniel Glossman-Mitnik" To: "Per-Ola Norrby" , References: <002a01c35d09$9d6a63f0$8400000a- at -laquicom02> Subject: a new question on comparing structures Date: Fri, 8 Aug 2003 10:56:56 -0500 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_00B5_01C35D9B.C9197A30" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2800.1158 Disposition-Notification-To: "Dr. Daniel Glossman-Mitnik" X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1165 This is a multi-part message in MIME format. ------=_NextPart_000_00B5_01C35D9B.C9197A30 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Re: CCL:summary for program for superimposing (small) Dear Per-Ola Norrby: I understand your point. But I really don't want to compare wit X-rays, = but to do a comparison between calculated structures. Suppose, for example, that I optimize = the geometry of a molecule=20 with increasing basis set levels. Thus I have an STO-3G, 3-21G, 6-31G, = 6-311G optimized structures.=20 Then I calculate the RMSD between each of the structures and that coming = > from the 6-311G=20 optimization (the highest level in this example). I guess that the RMSD = is going to decrease from=20 a maximum value for the STO-3G down to 0 corresponding to the 6-311G = with itself. Of course, the best match will be 0. But beside this, can we take a RMSD = value (what I call the threshold value in my previous message) and consider it closely enough = to the best match ? In order words, that the structure optimized with, say 6-31G, will be = almost as good as 6-311G ? I know that the example I has given is a very simple and moreover = abstract one. That is why I asked my question in the CCL hoping that the experience of the members = with different molecular geometry optimizations will help me. Of course, I understand = the problem in terms of listings of bonds and angles. Thanks in advance Daniel ********************************************************************** Dr. Daniel Glossman-Mitnik Centro de Investigaci=F3n en Materiales Avanzados (CIMAV) LAQUICOM - Laboratorio de Qu=EDmica Computacional Miguel de Cervantes 120 - Comp. Ind. Chihuahua Chihuahua, Chih. 31109 - MEXICO Tel.: (52) 614 4391151 FAX: (52) 614 4391112 E-mail: daniel.glossman{at}cimav.edu.mx glossman{at}hotmail.com dglossman{at}yahoo.com ********************************************************************** ----- Original Message -----=20 From: Per-Ola Norrby=20 To: Dr. Daniel Glossman-Mitnik=20 Cc: chemistry{at}ccl.net=20 Sent: Friday, August 08, 2003 5:41 AM Subject: Re: CCL:summary for program for superimposing (small) = molecules and new question Dear Daniel, Before I present the summary, I was to ask the (neccessary) follow-up question: In your experience, when a comparison of an experimental molecular structure (i.e. X-rays) and its calculated (Gaussian) counterpart is done, which will be the treshold RMSD value which can be considered acceptable ? The RMSD value is NOT a good measure of quality. One reason = is that soft modes (mainly torsions) are easily distorted by crystal = packing, leading to huge discrepancies if you only look at RMSD. A case = in point is biphenyl, where the crystal adopts a planar structure = becuase it packs better, whereas the gas phase global minimum has a = torsion angle around 41 degrees. You can either use a comparison that is insensitive to these = discrepancies, or do the calculation in the crystal (requiring QM with = periodic boundaries, not in Gaussian). I'd recommend comparison of = lists of bond lengths and angles. Alternatively, you can use the energy = of the crystal structure calculated by Gaussian and compare it to the = global minimum in Gaussian, but that requires constrained relaxation of = the crystal structure to allow for differences in paradigm between X-ray = and QM. Best regards, Per-Ola Norrby --=20 Per-Ola Norrby, Assoc. Professor, http://compchem.dfh.dk/PeO/ Technical University of Denmark, Department of Chemistry Building 201, Kemitorvet, DK-2800 Kgs. Lyngby, Denmark Email: pon{at}kemi.dtu.dk tel +45-45252123, fax +45-45933968 ------=_NextPart_000_00B5_01C35D9B.C9197A30 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Re: CCL:summary for program for superimposing = (small)
 
Dear Per-Ola Norrby:
 
I understand your point. But I really = don't want to=20 compare wit X-rays, but to do a comparison
between calculated = structures. Suppose, for=20 example, that  I optimize the geometry of a molecule =
with increasing basis set levels. Thus I have an STO-3G, 3-21G, 6-31G, = 6-311G=20 optimized structures. 
Then I calculate the=20 RMSD between each of the structures and that coming from the 6-311G=20
optimization (the=20 highest level in this example). I guess that the RMSD is going to = decrease=20  from
a maximum value for=20 the STO-3G down to 0 corresponding to the 6-311G with = itself.
 
Of course, the best match will be 0. = But beside=20 this, can we take a RMSD value (what I call the
threshold value in my previous message) = and=20 consider it closely enough to the best match ?
In order words, that the structure = optimized with,=20 say 6-31G, will be almost as good as 6-311G ?
 
I know that the example I has given is = a very=20 simple and moreover abstract one. That is why I
asked my question in the CCL hoping = that the=20 experience of the members with different
molecular geometry optimizations will = help me. Of=20 course, I understand the problem in terms
of listings of bonds and = angles.
 
Thanks in advance
 
          &nbs= p;            = ;            =             &= nbsp;          =20 Daniel
 
********************************************************************= **
Dr.=20 Daniel Glossman-Mitnik
Centro de Investigaci=F3n en Materiales = Avanzados=20 (CIMAV)
LAQUICOM - Laboratorio de Qu=EDmica Computacional
Miguel = de=20 Cervantes 120 - Comp. Ind. Chihuahua
Chihuahua, Chih. 31109 - = MEXICO
Tel.:=20 (52) 614 4391151
FAX: (52) 614 4391112
E-mail: daniel.glossman{at}cimav.edu.mx=
           = ;=20 glossman{at}hotmail.com
 &n= bsp;         =20 dglossman{at}yahoo.com
**********= ************************************************************
----- Original Message -----
From:=20 Per-Ola = Norrby=20
Sent: Friday, August 08, 2003 = 5:41=20 AM
Subject: Re: CCL:summary for = program for=20 superimposing (small) molecules and new question

        Dear=20 Daniel,

Before I present=20 the summary, I was to ask the (neccessary)
follow-up=20 question:  In your experience, when a = comparison
of an = experimental=20 molecular structure (i.e. X-rays) and its
calculated=20 (Gaussian) counterpart is done, which will be = the
treshold RMSD value=20 which can be considered acceptable ?

        The = RMSD value=20 is NOT a good measure of quality.  One reason is that soft modes = (mainly=20 torsions) are easily distorted by crystal packing, leading to huge=20 discrepancies if you only look at RMSD.  A case in point is = biphenyl,=20 where the crystal adopts a planar structure becuase it packs better, = whereas=20 the gas phase global minimum has a torsion angle around 41 = degrees.

        You can = either=20 use a comparison that is insensitive to these discrepancies, or do the = calculation in the crystal (requiring QM with periodic boundaries, not = in=20 Gaussian).  I'd recommend comparison of lists of bond lengths and = angles.  Alternatively, you can use the energy of the crystal = structure=20 calculated by Gaussian and compare it to the global minimum in = Gaussian, but=20 that requires constrained relaxation of the crystal structure to allow = for=20 differences in paradigm between X-ray and QM.

        Best=20 regards,

        Per-Ola = Norrby
--=20
Per-Ola Norrby, Assoc. Professor,=20 http://compchem.dfh.dk/PeO/
Technical University of Denmark, = Department of=20 Chemistry
Building 201, Kemitorvet, DK-2800 Kgs. Lyngby, = Denmark
Email:=20 pon{at}kemi.dtu.dk  tel +45-45252123,  fax=20 +45-45933968
------=_NextPart_000_00B5_01C35D9B.C9197A30--