From owner-chemistry@ccl.net Thu Jun 2 01:24:01 2011 From: "Kalju Kahn kalju{}chem.ucsb.edu" To: CCL Subject: CCL: CBS extrapolation formulas Message-Id: <-44812-110601172901-28431-KFVCTcC3yMw1iONrmBIlNw===server.ccl.net> X-Original-From: "Kalju Kahn" Content-Transfer-Encoding: 8bit Content-Type: text/plain;charset=iso-8859-1 Date: Wed, 1 Jun 2011 14:22:27 -0700 MIME-Version: 1.0 Sent to CCL by: "Kalju Kahn" [kalju(~)chem.ucsb.edu] Bradley, The answer depends on the level of theory. The SCF energy in basis sets that are fully radially saturated decreases exponentially with the maximum angular momentum. Dunning's basis of course are not fully radially saturated, but the exponential convergence seems to hold up well in practice. In my experience with mainly small stable molecules (alkanes, alcohols, ethers ...), exponential fitting of DZ to 6Z data yields all right fits for the SCF energy. You can see some examples and further discussion in "Improved efficiency of focal point conformational analysis with truncated correlation consistent basis sets" in J. Comp. Chem. 29(6), 900–911, (2008). Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20848/abstract. However, I would recommend a three-point extrapolation with {QZ, 5Z, 6Z} if you have up to 6Z data for two reasons. First, one can argue that the DZ and TZ data points just contaminate the better data. Second, with three points you can apply the analytical formula for E_inf rather than worry about the nonlinear regression. The analytical formula can be derived as: Solve[{e3 == Ei + A*Exp[-3*b], e4 == Ei + A*Exp[-4*b], e5 == Ei + A*Exp[-5*b]},{Ei, A, b}] Ei -> (e3*e5 - e4*e4)/(e3 - 2e4 + e5) The situation is more complex at the correlated level. In general, the SCF energy and the correlation energy should be extrapolated separately. The correlation energy converges as some sort of inverse power law. But pair correlations come in singlet pairs and triplet pairs; these converge differently. For example, the second order contribution (MP2-SCF) due to triplet pairs converges rather fast according to formula like: E_x(t2) = E_inf(t2) + At(X+0.5)^(-5) E(2) due to singlet pairs converges much slower, something like: E_x(s2) = E_inf(s2) + As(X+0.5)^(-3) Here X is the cardinal number and (X+0.5) is used as a compromise for molecules made up of hydrogen and heavy atoms. Use X+1 for systems made of heavy atoms only. Note that you can also derive an analytical expression for E_inf(2) and use two of the best energies to calculate the basis set limit. *** Do not use DZ data in second order energy extrapolation. If you do, the result will differ significantly from the true limit *** As you can imagine, things get messier at higher orders of perturbation theory. The convergence of the third order contributions (E(3)=MP3-MP2) can be described fairly well by: Triplet pairs: E_x(t3) = E_inf(t3) + At(X+0.5)^(-3) + Bt(X+0.5)^(-6) Singlet pairs: E_x(s3) = E_inf(s3) + As(X+0.5)^(-3) + Bs(X+0.5)^(-4) However, for this extrapolation, one shall not use DZ nor TZ data as they defy the converging trend. Please see "Convergence of third order correlation energy in atoms and molecules" in J Comp Chem, 28, (2), 547–554 (2006) and references there for details. Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20562/abstract I am not terribly well qualified to speak about the CC energy convergence, but I would expect an inverse power series convergence of correlation energy with the leading term of A(X+0.5)^(-3). I would expect that it is best to exclude DZ and TZ data points when extrapolating CC energies; especially when using non-augmented cc-pVXZ basis sets. Some relativistic corrections seem to have horrible basis set convergence (as if X^(-3) for the correlation energy was not horrible enough). You cannot blindly apply these formulas for geometries and properties. For example, nothing good comes out when you try to extrapolate the bond length or the fundamental frequency of H-O stretch in water based on cc-pVXZ data using {DZ, TZ, QZ} values. Please see "Anharmonic vibrational analysis of water with traditional and explicitly correlated coupled cluster methods" in J. Chem. Phys. 133, 074106 (2010) and references there for details. Link: http://jcp.aip.org/resource/1/jcpsa6/v133/i7/p074106_s1 Finally, while answering your question allowed me to sprinkle in couple of references that I am personally responsible for, please note that there is a significant body of theoretical and empirical work in this area done by too many people to name here; hopefully you will find the key references > from the three papers that I have cited above. Best luck, Kalju > > Sent to CCL by: "Bradley Kenneth Welch" [bwelch5=slu.edu] > Hey CCL readers, > > We want to do the CBS for a set of energies done with the dunning basis > sets from double zeta to 6 zeta. What are some of the commonly used > formulas for this? Which ones allow for 5 energies to be used in > extrapolation? > > > Sincerely, > > Bradley Welch > Saint Louis University > Room 218> > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Dr. Kalju Kahn Department of Chemistry and Biochemistry UC Santa Barbara, CA 93106 From owner-chemistry@ccl.net Thu Jun 2 09:54:01 2011 From: "Herbert Fruchtl herbert.fruchtl!A!st-andrews.ac.uk" To: CCL Subject: CCL:G: How to build a linux cluster for computational chemistry purposes?, HELP Message-Id: <-44813-110602053208-1127-iMsFkmoTzocQYFvDluq9eA===server.ccl.net> X-Original-From: Herbert Fruchtl Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=ISO-8859-1; format=flowed Date: Thu, 02 Jun 2011 10:31:41 +0100 MIME-Version: 1.0 Sent to CCL by: Herbert Fruchtl [herbert.fruchtl=st-andrews.ac.uk] Gaussian should work on pretty much any distribution. It depends more on what else you want to have on the front-end (I had problems with Ubuntu and GaussView, for example). I would go for the server oriented distributions that come without too much clutter by default (e.g. CentOS). You will need a queueing system. SGE (or whatever it's called now after the demise of Sun and the free version being abandoned by Oracle) installs out of the box and is probably the most administrator friendly. You can also use a cluster distribution like ROCKS. If you don't want to tweak it too much, you can have a fully running cluster in a day. If you do want things differently, it's quite a learning curve, because they do everything in a different way from standard Linux distros. HTH, Herbert On 06/01/2011 11:15 PM, Miguel Quiliano Meza rifaximina|,|gmail.com wrote: > Dear community. > > After... talk.. and talk.. with my boss, I convinced him to build a linux > cluster for computational chemistry purposes, in other words... to use > Gaussian03. So now I have 80 cores at my disposal. We also have TCP/LINDA. > > Could you please recommend me, one operative system and some online guide to > build it or book? Personal experiences are welcome. > > Thanks in advance. > > Miguel Quiliano. -- Herbert Fruchtl Senior Scientific Computing Officer School of Chemistry, School of Mathematics and Statistics University of St Andrews -- The University of St Andrews is a charity registered in Scotland: No SC013532 From owner-chemistry@ccl.net Thu Jun 2 10:29:01 2011 From: "Kalju Kahn kalju**chem.ucsb.edu" To: CCL Subject: CCL: CBS extrapolation formulas Message-Id: <-44814-110601172824-25361-1M5E3iT08RCUhza1A/Rirw#,#server.ccl.net> X-Original-From: "Kalju Kahn" Content-Transfer-Encoding: 8bit Content-Type: text/plain;charset=iso-8859-1 Date: Wed, 1 Jun 2011 14:18:26 -0700 MIME-Version: 1.0 Sent to CCL by: "Kalju Kahn" [kalju : chem.ucsb.edu] Bradley, It depends on the level of theory. The SCF energy in basis sets that are fully radially saturated decreases exponentially with the maximum angular momentum. Dunning's basis of course are not fully radially saturated, but the exponential convergence seems to hold up well in practice. In my experience with mainly small stable molecules (alkanes, alcohols, ethers ...), exponential fitting of DZ to 6Z data yields all right fits. You can see some examples and further discussion in "Improved efficiency of focal point conformational analysis with truncated correlation consistent basis sets" in J. Comp. Chem. 29(6), 900–911, (2008). Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20848/abstract. However, I would recommend a three-point extrapolation with {QZ, 5Z, 6Z} if you have up to 6Z data for two reasons. First, one can argue that the DZ and TZ data points just contaminate the better data. Second, with three points you can apply the analytical formula for E_inf rather than worry about the nonlinear regression. The analytical formula can be derived as: Solve[{e3 == Ei + A*Exp[-3*b], e4 == Ei + A*Exp[-4*b], e5 == Ei + A*Exp[-5*b]},{Ei, A, b}] Ei -> (e3*e5 - e4*e4)/(e3 - 2e4 + e5) The situation is more complex at the correlated level. In general, the SCF energy and the correlation energy should be extrapolated separately. The correlation energy converges as some sort of inverse power law. But pair correlations come in singlet pairs and triplet pairs; these converge differently. For example, at the second order contribution (MP2-SCF) level, triplet pairs converge rather fast according to formula like: E_x(t2) = E_inf(t2) + At(X+0.5)^(-5) Singlet pairs converge slower, something like: E_x(s2) = E_inf(s2) + As(X+0.5)^(-3) Here X is the cardinal number and (X+0.5) is used as a compromise for molecules made up of hydrogen and heavy atoms. Use X+1 for systems made of heavy atoms only. Note that you can also derive an analytical expression for E_inf(2) and use two of the best energies to calculate the basis set limit. *** Do not use DZ data in second order energy extrapolation. If you do, the result will differ significantly from the true limit *** As you can imagine, things get messier at higher orders of perturbation theory. The convergence of third order contributions (MP3-MP2) can be described fairly well by: Triplet pairs: E_x(t3) = E_inf(t3) + At(X+0.5)^(-3) + Bt(X+0.5)^(-6) Singlet pairs: E_x(s3) = E_inf(s3) + As(X+0.5)^(-3) + Bt(X+0.5)^(-4) However, for this extrapolation, one shall not use DZ nor TZ data as they defy the converging trend. Please see "Convergence of third order correlation energy in atoms and molecules" in J Comp Chem, 28, (2), 547–554 (2006) and references there for details. Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20562/abstract I am not terribly well qualified to speak about the CC energy convergence, but I would expect an inverse power series convergence of correlation energy with the leading term of A(X+0.5)^(-3). I would expect that it is best to exclude DZ and TZ data points when extrapolating CC energies; especially when using non-augmented cc-pVXZ basis sets. Some relativistic corrections seem to have horrible basis set convergence (as if X^(-3) for the correlation energy was not horrible enough). You cannot blindly apply these formulas for geometries and properties. For example, nothing good comes out when you try to extrapolate the bond length or the fundamental frequency of H-O stretch in water based on cc-pVXZ data using {DZ, TZ, QZ} values. Please see "Anharmonic vibrational analysis of water with traditional and explicitly correlated coupled cluster methods" in J. Chem. Phys. 133, 074106 (2010) and references there for details. Link: http://jcp.aip.org/resource/1/jcpsa6/v133/i7/p074106_s1 Finally, while answering your question allowed me to sprinkle in couple of references that I am personally responsible for, please note that there is a significant body of theoretical and empirical work in this area done by too many people to name here; hopefully you will find the key references > from the three papers that I have cited above. Best luck, Kalju >> Sent to CCL by: "Bradley Kenneth Welch" [bwelch5=slu.edu] >> Hey CCL readers, >> >> We want to do the CBS for a set of energies done with the dunning basis >> sets from double zeta to 6 zeta. What are some of the commonly used >> formulas for this? Which ones allow for 5 energies to be used in >> extrapolation? >> >> >> Sincerely, >> >> Bradley Welch >> Saint Louis University >> Room 218>> >> >> > > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > Dr. Kalju Kahn > Department of Chemistry and Biochemistry > UC Santa Barbara, CA 93106 > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Dr. Kalju Kahn Department of Chemistry and Biochemistry UC Santa Barbara, CA 93106 From owner-chemistry@ccl.net Thu Jun 2 11:04:01 2011 From: "harish sankar hariaimar..gmail.com" To: CCL Subject: CCL:G: How to build a linux cluster for computational chemistry purposes?, HELP Message-Id: <-44815-110602021714-10226-SE5vIe2fZTIwunGaKuFXgA^^^server.ccl.net> X-Original-From: harish sankar Content-Type: multipart/alternative; boundary=000e0cd605301f630804a4b49611 Date: Thu, 2 Jun 2011 11:47:06 +0530 MIME-Version: 1.0 Sent to CCL by: harish sankar [hariaimar{}gmail.com] --000e0cd605301f630804a4b49611 Content-Type: text/plain; charset=ISO-8859-1 :) :) :) 80 crore, anyways try the linux version pelican..it supports parallel computing.. On Thu, Jun 2, 2011 at 3:45 AM, Miguel Quiliano Meza rifaximina|,|gmail.com wrote: > Dear community. > > After... talk.. and talk.. with my boss, I convinced him to build a linux > cluster for computational chemistry purposes, in other words... to use > Gaussian03. So now I have 80 cores at my disposal. We also have TCP/LINDA. > > Could you please recommend me, one operative system and some online guide > to build it or book? Personal experiences are welcome. > > Thanks in advance. > > Miguel Quiliano. > --000e0cd605301f630804a4b49611 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable :) :) :) 80 crore, anyways try the linux version pelican..it supports paral= lel computing..

On Thu, Jun 2, 2011 at 3:= 45 AM, Miguel Quiliano Meza rifaximina|,|gmail= .com <o= wner-chemistry/./ccl.net> wrote:
Dear community.

After... talk.. and = talk.. with my boss, I convinced him= to build a linux cluster for computational chemistry purposes, in other wo= rds... to use Gaussian03. So now I have 80 cores at my disposal. W= e also have TCP/LINDA.

Could you please recommend me, one operative system and some online gui= de to build it or book? Personal experiences are welcome.

Thanks in = advance.

Miguel Quiliano.

--000e0cd605301f630804a4b49611-- From owner-chemistry@ccl.net Thu Jun 2 11:39:00 2011 From: "David Sherrill sherrill^-^gatech.edu" To: CCL Subject: CCL: Grad Student Awards in Comp Phys Chem Message-Id: <-44816-110602110218-6649-sSf71vWAguoQUUQXXAHEzw__server.ccl.net> X-Original-From: David Sherrill Content-Type: TEXT/PLAIN; format=flowed; charset=US-ASCII Date: Thu, 2 Jun 2011 11:01:05 -0400 (EDT) MIME-Version: 1.0 Sent to CCL by: David Sherrill [sherrill- -gatech.edu] Dear colleagues: The American Chemical Society subdivision of Theoretical Chemistry is pleased to announce two Awards in Computational Physical Chemistry that are open to current graduate students, sponsored by NCSA. The competition is open to any graduate student (regardless of citizenship) who began graduate study after August 1, 2007, and who is an ACS member (or whose advisor is an ACS member). The awards applications should be sent by e-mail to the Subdivision Secretary, Jan Steckel [steckelj=-=pitt.edu] with the applicant's name and "student award" in the subject line of the email. The deadline for applications is August 1, 2011. One award will be for up to 60,000 service units and the second award will be for up to 30,000 service units on NCSA's Teragrid resources. Awardee selection will be made on a competitive basis. Applicants should be working on new and innovative computational chemistry methods or applications in theoretical chemistry. Applicants should prepare a written description of a computational chemistry research project that requires high-performance computing, with an explanation of the scientific importance of the project. Proposals need to include an estimate of the computing resources required. Applicants should explain how they plan to use the computer time. Two letters of recommendation, including one from the student's advisor, along with a vita and transcript, are required. The proposal, including the vita, should not exceed four double-spaced pages. In addition, a faculty person (typically the applicant's research advisor) responsible for the applicant's use of the NCSA Teragrid resources must be identified. -- C. David Sherrill, Ph.D. Chair, ACS Subdivision of Theoretical Chemistry School of Chemistry & Biochemistry School of Computational Science and Engineering Georgia Institute of Technology www.chemistry.gatech.edu/faculty/Sherrill/ sherrill=-=gatech.edu, tel: 404-894-4037, fax: 404-894-7452 From owner-chemistry@ccl.net Thu Jun 2 12:14:00 2011 From: "Bradley Welch bwelch5 : slu.edu" To: CCL Subject: CCL: CBS extrapolation formulas Message-Id: <-44817-110602112020-9013-uewQPqof42KVb5rOou8ebA|server.ccl.net> X-Original-From: Bradley Welch Content-Type: multipart/alternative; boundary=00504502c71d4a3d8a04a4bc2cba Date: Thu, 2 Jun 2011 10:20:10 -0500 MIME-Version: 1.0 Sent to CCL by: Bradley Welch [bwelch5,slu.edu] --00504502c71d4a3d8a04a4bc2cba Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: quoted-printable Thank you for the help so far. Would it be legal to use an CBS extrapolation with energies gained from the correlated version of the SAPT program? And to go even further, would it be theoretically/chemically reasonable to do a CBS extrapolation of the SAPT calculation components (e_elec, e_ind, etc.) Bradley Welch On Wed, Jun 1, 2011 at 4:22 PM, Kalju Kahn kalju{}chem.ucsb.edu < owner-chemistry---ccl.net> wrote: > > Sent to CCL by: "Kalju Kahn" [kalju(~)chem.ucsb.edu] > Bradley, > > The answer depends on the level of theory. The SCF energy in basis sets > that are fully radially saturated decreases exponentially with the maximu= m > angular momentum. Dunning's basis of course are not fully radially > saturated, but the exponential convergence seems to hold up well in > practice. > > In my experience with mainly small stable molecules (alkanes, alcohols, > ethers ...), exponential fitting of DZ to 6Z data yields all right fits > for the SCF energy. You can see some examples and further discussion in > "Improved efficiency of focal point conformational analysis with truncate= d > correlation consistent basis sets" in J. Comp. Chem. 29(6), 900=96911, > (2008). > Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20848/abstract. > > However, I would recommend a three-point extrapolation with {QZ, 5Z, 6Z} > if you have up to 6Z data for two reasons. First, one can argue that the > DZ and TZ data points just contaminate the better data. Second, with > three points you can apply the analytical formula for E_inf rather than > worry about the nonlinear regression. The analytical formula can be > derived as: > > Solve[{e3 =3D=3D Ei + A*Exp[-3*b], > e4 =3D=3D Ei + A*Exp[-4*b], > e5 =3D=3D Ei + A*Exp[-5*b]},{Ei, A, b}] > > Ei -> (e3*e5 - e4*e4)/(e3 - 2e4 + e5) > > The situation is more complex at the correlated level. In general, the > SCF energy and the correlation energy should be extrapolated separately. > The correlation energy converges as some sort of inverse power law. But > pair correlations come in singlet pairs and triplet pairs; these converge > differently. For example, the second order contribution (MP2-SCF) due to > triplet pairs converges rather fast according to formula like: > > E_x(t2) =3D E_inf(t2) + At(X+0.5)^(-5) > > E(2) due to singlet pairs converges much slower, something like: > > E_x(s2) =3D E_inf(s2) + As(X+0.5)^(-3) > > Here X is the cardinal number and (X+0.5) is used as a compromise for > molecules made up of hydrogen and heavy atoms. Use X+1 for systems made > of heavy atoms only. Note that you can also derive an analytical > expression for E_inf(2) and use two of the best energies to calculate the > basis set limit. > > *** Do not use DZ data in second order energy extrapolation. If you do, > the result will differ significantly from the true limit *** > > As you can imagine, things get messier at higher orders of perturbation > theory. The convergence of the third order contributions (E(3)=3DMP3-MP2= ) > can be described fairly well by: > > Triplet pairs: E_x(t3) =3D E_inf(t3) + At(X+0.5)^(-3) + Bt(X+0.5)^(-6) > Singlet pairs: E_x(s3) =3D E_inf(s3) + As(X+0.5)^(-3) + Bs(X+0.5)^(-4) > > However, for this extrapolation, one shall not use DZ nor TZ data as they > defy the converging trend. Please see "Convergence of third order > correlation energy in atoms and molecules" in J Comp Chem, 28, (2), > 547=96554 (2006) and references there for details. > Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20562/abstract > > I am not terribly well qualified to speak about the CC energy convergence= , > but I would expect an inverse power series convergence of correlation > energy with the leading term of A(X+0.5)^(-3). I would expect that it is > best to exclude DZ and TZ data points when extrapolating CC energies; > especially when using non-augmented cc-pVXZ basis sets. > > Some relativistic corrections seem to have horrible basis set convergence > (as if X^(-3) for the correlation energy was not horrible enough). > > You cannot blindly apply these formulas for geometries and properties. Fo= r > example, nothing good comes out when you try to extrapolate the bond > length or the fundamental frequency of H-O stretch in water based on > cc-pVXZ data using {DZ, TZ, QZ} values. Please see "Anharmonic > vibrational analysis of water with traditional and explicitly correlated > coupled cluster methods" in J. Chem. Phys. 133, 074106 (2010) and > references there for details. > Link: http://jcp.aip.org/resource/1/jcpsa6/v133/i7/p074106_s1 > > Finally, while answering your question allowed me to sprinkle in couple o= f > references that I am personally responsible for, please note that there i= s > a significant body of theoretical and empirical work in this area done by > too many people to name here; hopefully you will find the key references > > from the three papers that I have cited above. > > Best luck, > > Kalju > > > > > Sent to CCL by: "Bradley Kenneth Welch" [bwelch5=3Dslu.edu] > > Hey CCL readers, > > > > We want to do the CBS for a set of energies done with the dunning basis > > sets from double zeta to 6 zeta. What are some of the commonly used > > formulas for this? Which ones allow for 5 energies to be used in > > extrapolation? > > > > > > Sincerely, > > > > Bradley Welch > > Saint Louis University > > Room 218> > > > > > > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > Dr. Kalju Kahn > Department of Chemistry and Biochemistry > UC Santa Barbara, CA 93106 > > > > -=3D This is automatically added to each message by the mailing script = =3D-> > > --00504502c71d4a3d8a04a4bc2cba Content-Type: text/html; charset=windows-1252 Content-Transfer-Encoding: quoted-printable Thank you for the help so far.

Would it be legal to use an CBS extr= apolation with energies gained from the correlated version of the SAPT prog= ram? And to go even further, would it be theoretically/chemically reasonabl= e to do a CBS extrapolation of the SAPT calculation components (e_elec, e_i= nd, etc.)

Bradley Welch

On Wed, Jun 1, 2011 at= 4:22 PM, Kalju Kahn kalju{}chem.ucsb.edu<= /a> <owner-= chemistry---ccl.net> wrote:

Sent to CCL by: "Kalju Kahn" [kalju(~)chem.ucsb.edu]
Bradley,

The answer depends on the level of theory. =A0The SCF energy in basis sets<= br> that are fully radially saturated decreases exponentially with the maximum<= br> angular momentum. =A0Dunning's basis of course are not fully radially saturated, but the exponential convergence seems to hold up well in
practice.

In my experience with mainly small stable molecules (alkanes, alcohols,
ethers ...), exponential fitting of DZ to 6Z data yields all right fits
for the SCF energy. You can see some examples and further discussion in
"Improved efficiency of focal point conformational analysis with trunc= ated
correlation consistent basis sets" in J. Comp. Chem. 29(6), 900=96911,=
(2008).
Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20848= /abstract.

However, I would recommend a three-point extrapolation with {QZ, 5Z, 6Z} if you have up to 6Z data for two reasons. =A0First, one can argue that the=
DZ and TZ data points just contaminate the better data. =A0Second, with
three points you can apply the analytical formula for E_inf rather than
worry about the nonlinear regression. =A0The analytical formula can be
derived as:

Solve[{e3 =3D=3D Ei + A*Exp[-3*b],
=A0 =A0 =A0 e4 =3D=3D Ei + A*Exp[-4*b],
=A0 =A0 =A0 e5 =3D=3D Ei + A*Exp[-5*b]},{Ei, A, b}]

Ei -> (e3*e5 - e4*e4)/(e3 - 2e4 + e5)

The situation is more complex at the correlated level. =A0In general, the SCF energy and the correlation energy should be extrapolated separately. The correlation energy converges as some sort of inverse power law. =A0But<= br> pair correlations come in singlet pairs and triplet pairs; these converge differently. For example, the second order contribution (MP2-SCF) due to triplet pairs converges rather fast according to formula like:

E_x(t2) =3D E_inf(t2) + At(X+0.5)^(-5)

E(2) due to singlet pairs converges much slower, something like:

E_x(s2) =3D E_inf(s2) + As(X+0.5)^(-3)

Here X is the cardinal number and (X+0.5) is used as a compromise for
molecules made up of hydrogen and heavy atoms. =A0Use X+1 for systems made<= br> of heavy atoms only. =A0Note that you can also derive an analytical
expression for E_inf(2) and use two of the best energies to calculate the basis set limit.

*** Do not use DZ data in second order energy extrapolation. =A0If you do,<= br> the result will differ significantly from the true limit ***

As you can imagine, things get messier at higher orders of perturbation
theory. =A0The convergence of the third order contributions (E(3)=3DMP3-MP2= )
can be described fairly well by:

Triplet pairs: E_x(t3) =3D E_inf(t3) + At(X+0.5)^(-3) + Bt(X+0.5)^(-6)
Singlet pairs: E_x(s3) =3D E_inf(s3) + As(X+0.5)^(-3) + Bs(X+0.5)^(-4)

However, for this extrapolation, one shall not use DZ nor TZ data as they defy the converging trend. =A0Please see "Convergence of third order correlation energy in atoms and molecules" in J Comp Chem, 28, (2), 547=96554 (2006) and references there for details.
Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20562= /abstract

I am not terribly well qualified to speak about the CC energy convergence,<= br> but I would expect an inverse power series convergence of correlation
energy with the leading term of A(X+0.5)^(-3). =A0I would expect that it is=
best to exclude DZ and TZ data points when extrapolating CC energies;
especially when using non-augmented cc-pVXZ basis sets.

Some relativistic corrections seem to have horrible basis set convergence (as if X^(-3) for the correlation energy was not horrible enough).

You cannot blindly apply these formulas for geometries and properties. For<= br> example, nothing good comes out when you try to extrapolate the bond
length or the fundamental frequency of H-O stretch in water based on
cc-pVXZ data using {DZ, TZ, QZ} values. =A0Please see "Anharmonic
vibrational analysis of water with traditional and explicitly correlated coupled cluster methods" =A0in J. Chem. Phys. 133, 074106 (2010) and references there for details.
Link: http://jcp.aip.org/resource/1/jcpsa6/v133/i7/p074106_s1

Finally, while answering your question allowed me to sprinkle in couple of<= br> references that I am personally responsible for, please note that there is<= br> a significant body of theoretical and empirical work in this area done by too many people to name here; hopefully you will find the key references > from the three papers that I have cited above.

Best luck,

Kalju

>
> Sent to CCL by: "Bradley Kenneth Welch" [bwelch5=3Dslu.edu]
> Hey CCL readers,
>
> We want to do the CBS for a set of energies done with the dunning basi= s
> sets from double zeta to 6 zeta. What are some of the commonly used > formulas for this? Which ones allow for 5 energies to be used in
> extrapolation?
>
>
> Sincerely,
>
> Bradley Welch
> Saint Louis University
> Room 218>
>
>


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr. Kalju Kahn
Department of Chemistry and Biochemistry
UC Santa Barbara, CA 93106

--00504502c71d4a3d8a04a4bc2cba-- From owner-chemistry@ccl.net Thu Jun 2 12:49:00 2011 From: "Jorl Jose Montalvo montalvo161-,-hotmail.com" To: CCL Subject: CCL:G: How to install gaussian 09 in Open suse 11.4? Message-Id: <-44818-110602104804-11803-taquGaRCuZgQZdOp2wtNPA{}server.ccl.net> X-Original-From: "Jorl Jose Montalvo" Date: Thu, 2 Jun 2011 10:48:02 -0400 Sent to CCL by: "Jorl Jose Montalvo" [montalvo161 . hotmail.com] Dear community. I have tried to install gaussian 09 on suse 11.4 in accordance with the instructions provided by the G09's main page but I could not do exactly .... I had problems with the files .login and .profile, which does not encounter in open suse. Could someone tell me an G09 installation manual for suse 11.4? or where files (.login and .profile) are located in suse 11.4? From owner-chemistry@ccl.net Thu Jun 2 13:24:01 2011 From: "Joel Jos Montalvo montalvo161#hotmail.com" To: CCL Subject: CCL:G: How to install gaussian 09 in open suse 11.4? Message-Id: <-44819-110602125409-1435-wYvFjSzW+By7szwHaIgRWg[*]server.ccl.net> X-Original-From: "Joel Jos Montalvo" Date: Thu, 2 Jun 2011 12:54:06 -0400 Sent to CCL by: "Joel Jos Montalvo" [montalvo161-#-hotmail.com] Hi everyone, I tried to install gaussian 09 at open suse 11.4 according to the manual provided by the official website of gaussian, but I had problems with the files .login and .profile, since they are not found in this distribution. could someone tell me how to install G09 in suse 11.4? or where I find the files .login and .profile in this distribution? From owner-chemistry@ccl.net Thu Jun 2 13:59:01 2011 From: "jaleel uc jaleel.uc#%#gmail.com" To: CCL Subject: CCL:G: How to build a linux cluster for computational chemistry purposes?, HELP Message-Id: <-44820-110602124839-27848-/fsl0ai0OwPl1b9TgENvhA(-)server.ccl.net> X-Original-From: jaleel uc Content-Type: multipart/alternative; boundary=0016e6d63ffa4a45d304a4bd68a9 Date: Thu, 2 Jun 2011 22:18:32 +0530 MIME-Version: 1.0 Sent to CCL by: jaleel uc [jaleel.uc^gmail.com] --0016e6d63ffa4a45d304a4bd68a9 Content-Type: text/plain; charset=ISO-8859-1 dear friend this issue already discussed several time in this forum. please refer ccl archives for getting more idea . If my memory is correct please refer 2006-2007 regards jaleel On Thu, Jun 2, 2011 at 11:47 AM, harish sankar hariaimar..gmail.com < owner-chemistry[-]ccl.net> wrote: > :) :) :) 80 crore, anyways try the linux version pelican..it supports > parallel computing.. > > On Thu, Jun 2, 2011 at 3:45 AM, Miguel Quiliano Meza rifaximina|,| > gmail.com wrote: > >> Dear community. >> >> After... talk.. and talk.. with my boss, I convinced him to build a linux >> cluster for computational chemistry purposes, in other words... to use >> Gaussian03. So now I have 80 cores at my disposal. We also have >> TCP/LINDA. >> >> Could you please recommend me, one operative system and some online guide >> to build it or book? Personal experiences are welcome. >> >> Thanks in advance. >> >> Miguel Quiliano. >> > > -- Dr UCA JALEEL Asst Professor in cheminformatics UGC AIDED Centre for cheminformatics MCC calicut phone --0016e6d63ffa4a45d304a4bd68a9 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable =A0=A0 dear friend

=A0=A0 this issue already discussed several time= in this=A0 forum. please refer ccl archives for getting=A0 more idea . If = my memory is correct please refer 2006-2007

regards


jale= el

On Thu, Jun 2, 2011 at 11:47 AM, harish sankar h= ariaimar..gmail.com <= owner-chemistry[-]ccl.net><= /span> wrote:
:) :) :) 80 crore, anyways try the linux ve= rsion pelican..it supports parallel computing..

On Thu, Jun 2, 2011 at 3:45 AM, Miguel Quiliano Meza rifaximina|,|gmail.com <= ;owner-che= mistry ~ ccl.net> wrote:
Dear community.

After... talk.. and t= alk.. with my boss, I convinced him = to build a linux cluster for computational chemistry purposes, in other wor= ds... to use Gaussian03. So now I have 80 cores at my disposal. We= also have TCP/LINDA.

Could you please recommend me, one operative system and some online gui= de to build it or book? Personal experiences are welcome.

Thanks in = advance.

Miguel Quiliano.




--
=A0Dr UCA JALEEL
=A0= Asst Professor in cheminformatics
=A0UGC AIDED Centre for cheminformati= cs
=A0MCC calicut
phone

--0016e6d63ffa4a45d304a4bd68a9-- From owner-chemistry@ccl.net Thu Jun 2 14:33:00 2011 From: "PONNUCHAMY VEERA PANDIAN veera.pandi33 . gmail.com" To: CCL Subject: CCL: NBO doubt Message-Id: <-44821-110602092853-10079-2HvhIEUMNt6hhT6fjhJV6w__server.ccl.net> X-Original-From: "PONNUCHAMY VEERA PANDIAN" Date: Thu, 2 Jun 2011 09:28:51 -0400 Sent to CCL by: "PONNUCHAMY VEERA PANDIAN" [veera.pandi33{}gmail.com] Dear CCLers, I have done the NBO calculation on a -conjugated system and I got the following output. It gives the Second Order Perturbation Theory Analysis results. 126. LP ( 1) N 10 /125. LP*( 1) C 7 133.19 0.13 0.134 127. LP ( 1) N 11 /125. LP*( 1) C 7 131.71 0.13 0.134 129. LP ( 2) N 12 /836. BD*( 2) N 13 - N 14 96.18 0.17 0.117 In this I observed the interaction between LP(1) N 10 and LP*(1) C 7 will be the key. But I dont know what is LP*(1) C7 and how one can interpret this results. I expect your valuable suggestions. Thanks in advance... Your Sincerely, P.Veera From owner-chemistry@ccl.net Thu Jun 2 15:09:00 2011 From: "lukman olasunkanmi waleolasunkanmi%x%gmail.com" To: CCL Subject: CCL: thermodynamic parameters! Message-Id: <-44822-110602134534-25649-zDU53HmRoc2DGpL46zK8/A*o*server.ccl.net> X-Original-From: "lukman olasunkanmi" Date: Thu, 2 Jun 2011 13:45:31 -0400 Sent to CCL by: "lukman olasunkanmi" [waleolasunkanmi%%gmail.com] Hello, Please, can anyone explain why the heat of formation and change in free energy of formation of some simple known compounds like ethene, ethyne, benzene(l,g), HCN, N2O etc are positive and yet they are stable. From owner-chemistry@ccl.net Thu Jun 2 16:17:00 2011 From: "Carlos Silva Lopez carlos.silva#%#uvigo.es" To: CCL Subject: CCL: Last call for SimOC 2011. WATOC 2011 Satellite Meeting in Vigo Message-Id: <-44823-110602134303-23192-PfZ1xEZNe/RXbdly124K3w-,-server.ccl.net> X-Original-From: "Carlos Silva Lopez" Date: Thu, 2 Jun 2011 13:43:00 -0400 Sent to CCL by: "Carlos Silva Lopez" [carlos.silva]_[uvigo.es] Simulations in Organic Chemistry (SimOC 2011) Vigo (Spain) July 15-16th, University of Vigo. Chairs: Armando Navarro and Carlos Silva Website: http://webs.uvigo.es/simoc2011 This WATOC 2011 satellite meeting will gather lectures by 10 invited speakers who have made prominent contributions to the broad field of Computational Chemistry applied to Organic Chemistry challenges. The meeting also seeks to provide an excellent environment to interact with different researchers in the field, to promote technology transfer, and to establish new collaborations among the participants. To facilitate the participation of everyone, from professors to undergraduate students, we are glad to announce that the participation to SimOC 2011 is open to everyone at no cost. Registration is, however, necessary, since the number of participants is limited to 150 persons. Confirmed speakers: - David Birney (Texas Tech University) - Israel Fernndez (Universidad Complutense de Madrid) - Barry Carpenter (Cardiff University) - Olalla Nieto (Universidade de Vigo) - Ins Corral (Universidad Autonoma de Madrid) - Henry Rzepa (Imperial College London) - Dieter Cremer (Southern Methodist University) - Fabrizio Santoro (Istituto di Chimica dei Composti OrganoMetallici) - Ayan Datta (Indian Institute of Science Education and Research) - Miquel Sol (Universitat de Girona) - Duncan Poole (NVIDIA corporation) Vigo is conveniently located just 90 km (60 miles) south from Santiago de Compostela, where the WATOC conference will be held between July 17-22nd. Hourly trains connect both cities in 90 min trips. Vigo is a medium sized city with vibrant harbour activity, gorgeous marine landscapes and outdoor Celtic remains right at the center of the town. Carlos Silva and Armando Navarro Chairs of Simoc 2011 From owner-chemistry@ccl.net Thu Jun 2 16:52:00 2011 From: "Kalju Kahn kalju__chem.ucsb.edu" To: CCL Subject: CCL: CBS extrapolation formulas Message-Id: <-44824-110602160621-13730-tceN9kqW1nU1oSzAMbU9sA^server.ccl.net> X-Original-From: "Kalju Kahn" Content-Transfer-Encoding: 8bit Content-Type: text/plain;charset=iso-8859-1 Date: Thu, 2 Jun 2011 13:06:12 -0700 MIME-Version: 1.0 Sent to CCL by: "Kalju Kahn" [kalju ~ chem.ucsb.edu] Bradley, I have seen that being done (for example, look up the recent work of Pavel Hobza) and I do not see any obvious reasons why it would be not "legal". > From what I understand, interactions due to non-correlation terms should converge nearly exponentially, and the correlation-related terms, such as dispersion energy, converge more slowly (similar to or worse to E(2)). Again, non-augmented basis sets are not appropriate (one can surmise this for electrostatics from the finding that dipole moments do not converge smoothly in cc-pVXZ, see "Basis-set convergence of the molecular electric dipole moment" in JCP 11, (10), 4424 (1999)), and double-zeta results are best to be excluded when aug-cc-pVXZ data is used. You may want to evaluate if Frank Jensen's polarization consistent basis sets give the desired convergence of interaction energies faster and at lower cost. Best wishes, Kalju > Thank you for the help so far. > > Would it be legal to use an CBS extrapolation with energies gained from > the > correlated version of the SAPT program? And to go even further, would it > be > theoretically/chemically reasonable to do a CBS extrapolation of the SAPT > calculation components (e_elec, e_ind, etc.) > > Bradley Welch > > On Wed, Jun 1, 2011 at 4:22 PM, Kalju Kahn kalju{}chem.ucsb.edu < > owner-chemistry^ccl.net> wrote: > >> >> Sent to CCL by: "Kalju Kahn" [kalju(~)chem.ucsb.edu] >> Bradley, >> >> The answer depends on the level of theory. The SCF energy in basis sets >> that are fully radially saturated decreases exponentially with the >> maximum >> angular momentum. Dunning's basis of course are not fully radially >> saturated, but the exponential convergence seems to hold up well in >> practice. >> >> In my experience with mainly small stable molecules (alkanes, alcohols, >> ethers ...), exponential fitting of DZ to 6Z data yields all right fits >> for the SCF energy. You can see some examples and further discussion in >> "Improved efficiency of focal point conformational analysis with >> truncated >> correlation consistent basis sets" in J. Comp. Chem. 29(6), 900–911, >> (2008). >> Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20848/abstract. >> >> However, I would recommend a three-point extrapolation with {QZ, 5Z, 6Z} >> if you have up to 6Z data for two reasons. First, one can argue that >> the >> DZ and TZ data points just contaminate the better data. Second, with >> three points you can apply the analytical formula for E_inf rather than >> worry about the nonlinear regression. The analytical formula can be >> derived as: >> >> Solve[{e3 == Ei + A*Exp[-3*b], >> e4 == Ei + A*Exp[-4*b], >> e5 == Ei + A*Exp[-5*b]},{Ei, A, b}] >> >> Ei -> (e3*e5 - e4*e4)/(e3 - 2e4 + e5) >> >> The situation is more complex at the correlated level. In general, the >> SCF energy and the correlation energy should be extrapolated separately. >> The correlation energy converges as some sort of inverse power law. But >> pair correlations come in singlet pairs and triplet pairs; these >> converge >> differently. For example, the second order contribution (MP2-SCF) due to >> triplet pairs converges rather fast according to formula like: >> >> E_x(t2) = E_inf(t2) + At(X+0.5)^(-5) >> >> E(2) due to singlet pairs converges much slower, something like: >> >> E_x(s2) = E_inf(s2) + As(X+0.5)^(-3) >> >> Here X is the cardinal number and (X+0.5) is used as a compromise for >> molecules made up of hydrogen and heavy atoms. Use X+1 for systems made >> of heavy atoms only. Note that you can also derive an analytical >> expression for E_inf(2) and use two of the best energies to calculate >> the >> basis set limit. >> >> *** Do not use DZ data in second order energy extrapolation. If you do, >> the result will differ significantly from the true limit *** >> >> As you can imagine, things get messier at higher orders of perturbation >> theory. The convergence of the third order contributions (E(3)=MP3-MP2) >> can be described fairly well by: >> >> Triplet pairs: E_x(t3) = E_inf(t3) + At(X+0.5)^(-3) + Bt(X+0.5)^(-6) >> Singlet pairs: E_x(s3) = E_inf(s3) + As(X+0.5)^(-3) + Bs(X+0.5)^(-4) >> >> However, for this extrapolation, one shall not use DZ nor TZ data as >> they >> defy the converging trend. Please see "Convergence of third order >> correlation energy in atoms and molecules" in J Comp Chem, 28, (2), >> 547–554 (2006) and references there for details. >> Link: http://onlinelibrary.wiley.com/doi/10.1002/jcc.20562/abstract >> >> I am not terribly well qualified to speak about the CC energy >> convergence, >> but I would expect an inverse power series convergence of correlation >> energy with the leading term of A(X+0.5)^(-3). I would expect that it >> is >> best to exclude DZ and TZ data points when extrapolating CC energies; >> especially when using non-augmented cc-pVXZ basis sets. >> >> Some relativistic corrections seem to have horrible basis set >> convergence >> (as if X^(-3) for the correlation energy was not horrible enough). >> >> You cannot blindly apply these formulas for geometries and properties. >> For >> example, nothing good comes out when you try to extrapolate the bond >> length or the fundamental frequency of H-O stretch in water based on >> cc-pVXZ data using {DZ, TZ, QZ} values. Please see "Anharmonic >> vibrational analysis of water with traditional and explicitly correlated >> coupled cluster methods" in J. Chem. Phys. 133, 074106 (2010) and >> references there for details. >> Link: http://jcp.aip.org/resource/1/jcpsa6/v133/i7/p074106_s1 >> >> Finally, while answering your question allowed me to sprinkle in couple >> of >> references that I am personally responsible for, please note that there >> is >> a significant body of theoretical and empirical work in this area done >> by >> too many people to name here; hopefully you will find the key references >> > from the three papers that I have cited above. >> >> Best luck, >> >> Kalju >> >> > >> > Sent to CCL by: "Bradley Kenneth Welch" [bwelch5=slu.edu] >> > Hey CCL readers, >> > >> > We want to do the CBS for a set of energies done with the dunning >> basis >> > sets from double zeta to 6 zeta. What are some of the commonly used >> > formulas for this? Which ones allow for 5 energies to be used in >> > extrapolation? >> > >> > >> > Sincerely, >> > >> > Bradley Welch >> > Saint Louis University >> > Room 218> >> > >> > >> >> >> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ >> Dr. Kalju Kahn >> Department of Chemistry and Biochemistry >> UC Santa Barbara, CA 93106> >> >> > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Dr. Kalju Kahn Department of Chemistry and Biochemistry UC Santa Barbara, CA 93106 From owner-chemistry@ccl.net Thu Jun 2 17:34:00 2011 From: "Gustaf Olsson gustaf.olsson**lnu.se" To: CCL Subject: CCL:G: How to install gaussian 09 in open suse 11.4? Message-Id: <-44825-110602161116-17103-pjuh/hyP8WupL6/v5iqeiA * server.ccl.net> X-Original-From: Gustaf Olsson Content-Language: en-US Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="us-ascii" Date: Thu, 2 Jun 2011 22:11:01 +0200 MIME-Version: 1.0 Sent to CCL by: Gustaf Olsson [gustaf.olsson ~~ lnu.se] I checked the official site just to make sure since I'm not familiar with the gaussian installation process. But usually, this is the case here as well, the .profile (or .xxxxx files) is a local file on you computer, which usually needs to be modified to add the correct directories for executables to you $PATH. Run this command in the terminal cd $HOME (or cd ~) To get to you home directory. Then run ls -la You should have a lot of .XXXX (hidden files) files there. Hopefully one of these will be .login/.profile, this depends on the shell/system. Other alternatives that might be available can be .cshrc, .bashrc and so forth, one "profile" for each available shell! Best of luck to you On 2 jun 2011, at 18.54, Joel Jos Montalvo montalvo161#hotmail.com wrote: > > Sent to CCL by: "Joel Jos Montalvo" [montalvo161-#-hotmail.com] > Hi everyone, > I tried to install gaussian 09 at open suse 11.4 according to the manual provided by the official website of gaussian, but I had problems with the files .login and .profile, since they are not found in this distribution. > could someone tell me how to install G09 in suse 11.4? or where I find the files .login and .profile in this distribution?> > From owner-chemistry@ccl.net Thu Jun 2 23:14:00 2011 From: "neeraj misra neerajmisra%hotmail.com" To: CCL Subject: CCL:G: calculation Message-Id: <-44826-110602231221-8576-6lQRlmlMz9e7UmBL6YZJWg:+:server.ccl.net> X-Original-From: "neeraj misra" Date: Thu, 2 Jun 2011 23:12:18 -0400 Sent to CCL by: "neeraj misra" [neerajmisra]![hotmail.com] Dear members, The following is the input of a gaussian calculation and also the output.Even after running for three days it has possibly got stuck....any help to solve this would be appreciated. INPUT: # polar opt(TIGHT,diis) freq b3lyp/6-311++g(d,p)SCF(xqc,vshift=100,maxcycle=512) ABCD 0 1 Br 5.35450 -1.05260 -0.03250 O 2.31980 -1.65120 -0.04060 N 0.68670 2.69940 0.05660 N 1.51070 0.50300 0.00760 N 3.01230 2.38740 0.04830 N 5.34950 2.06700 0.03600 N -4.16190 -3.51580 -0.06830 N -6.02390 1.65680 0.04720 C 0.99300 -2.03300 -0.04620 C 0.33640 -2.19520 -1.25680 C 0.34110 -2.24720 1.15860 C -1.00410 -2.58080 -1.26250 C -0.99940 -2.63300 1.15290 C -1.67190 -2.79970 -0.05770 C 1.03450 -1.96540 -2.55670 C 1.04420 -2.07350 2.46460 C 2.58870 -0.30820 -0.01220 C 1.78820 1.82210 0.03710 C -0.70710 2.48280 0.05400 C 3.89120 0.14720 -0.00340 C 4.04410 1.51840 0.02740 C -1.55810 3.42210 -0.52860 C -1.23450 1.32970 0.63510 C -2.93650 3.20810 -0.53000 C -2.61280 1.11570 0.63360 C -3.46380 2.05500 0.05110 C -3.04530 -3.19470 -0.06350 C -4.87580 1.83540 0.04890 H -1.52990 -2.70860 -2.20660 H -1.52170 -2.80100 2.09260 H 0.47350 -2.39040 -3.39610 H 1.15330 -0.89280 -2.73970 H 2.02010 -2.44280 -2.56130 H 0.48460 -2.53110 3.28770 H 2.02820 -2.55370 2.44580 H 1.16710 -1.00980 2.69170 H 0.95950 3.67070 -0.05960 H -1.15940 4.32420 -0.98540 H -0.61420 0.60330 1.14640 H -3.58690 3.94930 -0.98830 H -3.01060 0.21830 1.10160 H 6.09980 1.51390 0.43120 H 5.42630 3.07600 0.06720 OUTPUT Requested convergence on RMS density matrix=1.00D-08 within 128 cycles. Requested convergence on MAX density matrix=1.00D-06. Requested convergence on energy=1.00D-06. Virtual orbitals will be shifted by 0.100 hartree. No special actions if energy rises.