From fritsch at hmi.de Thu Jan 6 11:53:08 2005 From: fritsch at hmi.de (Wolfgang Fritsch) Date: Thu, 06 Jan 2005 11:53:08 +0100 Subject: [CPMD-list] slow geometry optimization - would underrelaxation help? Message-ID: <41DD1894.1060704@hmi.de> Dear experts, I am trying to optimize the geometry of a CuGaSe2 cell. It seems to me that the convergence is awfully slow and hence wonder whether I am missing something. "Slow" means that with the 16 atoms in the cell, after a few hundreds of geometry steps convergence is not reached and actually, when I check the positions over optimization steps, it appears that they move for a while in one direction but then change into another, with no clear indication for final convergence yet. I guess I am missing something like an "under-relaxation factor" which would allow to change the coordinates of atoms only a given fraction of what happens now - leading to smaller steps but less jiggling forth and back. I remember the benevolent effect of such underrelaxation factors from codes in fluid dynamics. I append two sets of input. Maybe someone in the knowing sees my blunders right away. The first set is the one I am using now. The second set I had to abandon as it let the code stall: it runs on, say, 3 nodes with 4 cpus each but after a few minutes one of the cpus on each node would go idle and no further progress in the calculation would ensue (other cpus busy but no further output). I have tried another method, the steepest descend, but then progress in the optimization would be even slower. Thank you for your attention! Wolfgang Fritsch set 1: &CPMD OPTIMIZE GEOMETRY RESTART WAVEFUNCTION GEOFILE CONVERGENCE GEOMETRY 0.0001 CONVERGENCE ORBITALS 0.00005 CONVERGENCE CELL 1.0 PCG MINIMIZE TIMESTEP 3 MAXSTEP 40000 MEMORY BIG &END &DFT FUNCTIONAL BLYP &END &SYSTEM SYMMETRY 8 CUTOFF 110.0 CELL 10.6 1.0 1.961 0.0 0.0 0.0 &END &ATOMS *Cu_MT_BLYP_NLCC LMAX=D SKIP=P LOC=S 4 0.0 0.0 0.0 5.3101 0.0 5.1967 5.3101 5.3101 10.3935 0.0 5.3101 15.5902 *Ga_MT_GIA_BLYP.psp LMAX=D LOC=S 4 5.3101 5.3101 0.0 0.0 5.3101 5.1967 0.0 0.0 10.3935 5.3101 0.0 15.5902 *Se_MT_BLYP.psp LMAX=D LOC=D 8 2.6752 7.6524 2.6161 ...... CONSTRAINTS FIX ATOMS 1 1 END CONSTRAINTS &END ======================================================== set 2 &CPMD OPTIMIZE GEOMETRY RESTART WAVEFUNCTION GEOFILE CONVERGENCE GEOMETRY 0.0004 CONVERGENCE ORBITALS 0.00005 CONVERGENCE CELL 1.0 CONVERGENCE ADAPT 0.02 CONVERGENCE ENERGY 0.05 CONVERGENCE CALFOR 2.5 CONVERGENCE RELAX 5 PCG MINIMIZE TIMESTEP 3 MAXSTEP 40000 MEMORY BIG &END rest as above... From langel at uni-greifswald.de Thu Jan 6 11:53:02 2005 From: langel at uni-greifswald.de (Walter Langel) Date: Thu, 06 Jan 2005 11:53:02 +0100 Subject: [CPMD-list] "True" Eigenvalues In-Reply-To: References: <1101309574.41a4a6861f205@webmail.mail.gatech.edu> Message-ID: <41DD188E.6050307@uni-greifswald.de> Dear Ari, can you give me a reference for this "constant energy term dropped when eigenvalues are calculated". This probably also explains the shift of the eigenvalues as a function of cell size. Thank you Walter Langel Ari P Seitsonen wrote: >Dear Eung-Gun, > > I write what I know, some one please correct and add: In order to get the >correct eigenvalues you need to have the reference, that is, vacuum, >somewhere in your system. So you can have 2-dimensional (surface = slab >geometry), 1D (line) or 0D (isolated system). When you have performed your >calculation you check the value of the electro-static potential in the >(middle of) vacuum region, and subtract that from the eigenvalues. > > Greetings (again/still) from Zurich, > > apsi > >-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=- > Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/ > Tel +41 1 635 44 97 / Fax +41 1 635 68 38 / GSM +41 79 719 09 35 > Anschrift: Physikalisch Chemisches Institut (PCI), Universit?t Z?rich (UniZh) > Indirizzo: Winterthurerstrasse 190, CH-8057 Z?rich > Address: Schweiz / Svizzera / Suisse / Svizra / Switzerland > >On Wed, 24 Nov 2004, Eung-Gun Kim wrote: > > > >>Dear List Subscribers, >> >>I was told that in most plane-wave/pseudopotential codes, there is a constant >>energy term dropped when eigenvalues are calculated. For most purposes where >>only the energy difference matters, one would not need this dropped term. >>However, in >>some instances, for example, when one calculates the work function OR compares >>HOMO/LUMO levels between two different molecules, this dropped term is to be >>recovered. Is there a way to retrieve this missing term to obtain "true" >>eigenvalues? Thanks very much for reading. >> >>Best regards, >> >>EG Kim >> >>_______________________________________________ >>CPMD-list mailing list >>CPMD-list at cpmd.org >>http://cpmd.org/mailman/listinfo/cpmd-list >> >> >> >_______________________________________________ >CPMD-list mailing list >CPMD-list at cpmd.org >http://cpmd.org/mailman/listinfo/cpmd-list > > > -- Prof. Dr. Walter Langel Institut fuer Chemie und Biochemie Universitaet Greifswald Soldmannstrasse 23 D-17487 Greifswald Germany Tel +49 3834 86 4423 Fax +49 3834 86 4413 http://www.chemie.uni-greifswald.de/~plasma -------------- next part -------------- An HTML attachment was scrubbed... URL: http://cpmd.org/pipermail/cpmd-list/attachments/20050106/675b2811/attachment.html From ZRLSRB at ch.ibm.com Thu Jan 6 12:16:18 2005 From: ZRLSRB at ch.ibm.com (Salomon Billeter) Date: Thu, 6 Jan 2005 12:16:18 +0100 Subject: [CPMD-list] slow geometry optimization - would underrelaxation help? Message-ID: Dear Wolfgang, instead deliberately using too small a step, you can require your trial move to satisfy the two Wolfe conditions (sufficient decrease and curvature conditions). The L-BFGS geometry optimizer in CPMD (keyword LBFGS in the &CPMD section) has a built-in dynamic trust radius method for predicting maximum trial move lengths based on these conditions (and for rejecting steps in case they fail to reduce the energy). In addition, I'd recommend using adaptive wavefunction convergence criteria (CONVERGENCE ADAPT and CONVERGENCE INITIAL) for saving wavefunction cycles in earlier stages of the optimization. Best regards, Salomon From ZRLSRB at ch.ibm.com Thu Jan 6 12:27:04 2005 From: ZRLSRB at ch.ibm.com (Salomon Billeter) Date: Thu, 6 Jan 2005 12:27:04 +0100 Subject: [CPMD-list] "True" Eigenvalues Message-ID: Dear Walter, you might also want to have a look at http://cpmd.org/mailman/htdig/cpmd-list/2004-November/003808.html, which gives you the according terms in the CPMD source code. D. Marx and J. Hutter, "Ab-initio Molecular Dynamics: Theory and Implementation", Modern Methods and Algorithms in Quantum Chemistry, Forschungzentrum Juelich, NIC Series, vol. 1, (2000), contains also a good review about solving Poisson's equation in reciprocal space with Gaussian core charges. Best regards, Salomon From fritsch at hmi.de Fri Jan 7 12:24:24 2005 From: fritsch at hmi.de (Wolfgang Fritsch) Date: Fri, 07 Jan 2005 12:24:24 +0100 Subject: [CPMD-list] slow geometry optimization - would underrelaxation help? In-Reply-To: References: Message-ID: <41DE7168.8000306@hmi.de> Salomon Billeter wrote: > you can require your trial > move to satisfy the two Wolfe conditions (sufficient decrease and curvature > conditions). The L-BFGS geometry optimizer in CPMD (keyword LBFGS in the > &CPMD section) has a built-in dynamic trust radius method for predicting > maximum trial move lengths based on these conditions (and for rejecting > steps in case they fail to reduce the energy). Dear Salomon, many thanks for your hint. The geometry optimization now proceeds in remarkable ways. For a final comment I still have to await the resolution of the other problem which I already have mentioned, the hanging of the code because one of the cpus stalls. Don't know whether this is software or hardware related. I need to contact local support here. Thanks again, Wolfgang From claudio.melis at kcl.ac.uk Tue Jan 11 17:35:39 2005 From: claudio.melis at kcl.ac.uk (Claudio Melis) Date: 11 Jan 2005 16:35:39 +0000 Subject: [CPMD-list] Pseudopotential of Chlorine Message-ID: <1105461339.6339.12.camel@phpc32.ph.kcl.ac.uk> Dear CPMD users, I'm looking for a Martin Troullier pseudopotential with the PBE functional for chlorine. Does anybody know where I can find it ? Thanks Claudio Melis From fritsch at hmi.de Thu Jan 13 10:52:55 2005 From: fritsch at hmi.de (Wolfgang Fritsch) Date: Thu, 13 Jan 2005 10:52:55 +0100 Subject: [CPMD-list] error on cell optimization: "need some programming" Message-ID: <41E644F7.6080407@hmi.de> Hi experts, there are already quite a number of entries in the list for the topic of cell optimization. I haven't found anything for my problem, so please bear with me. I am trying to optimize the cell parameters for my case of CuGaSe2. After having, seemingly successfully, optimized the geometry first I then add the keywords CONVERGENCE CELL 0.05 STEEPEST DESCENT CELL and optionally ISOTROPIC CELL But the code stops with error messages after some wavefunction optimization. This is with the mpi version of the code on an alpha cluster. Any suggestion would be appreciated. I append the error message and then the input list. Thank you for your attention, Wolfgang ================================================================== THIS NEEDS FIRST SOME PROGRAMING ! THIS NEEDS FIRST SOME PROGRAMING ! PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 2] THIS NEEDS FIRST SOME PROGRAMING ! THIS NEEDS FIRST SOME PROGRAMING ! PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 0] THIS NEEDS FIRST SOME PROGRAMING ! PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 4] THIS NEEDS FIRST SOME PROGRAMING ! PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 10] THIS NEEDS FIRST SOME PROGRAMING ! ...and a few more of the kind, differreing by the number after PROC= ====================================================================== &CPMD OPTIMIZE GEOMETRY CONVERGENCE GEOMETRY 0.0005 CONVERGENCE ORBITALS 0.0005 CONVERGENCE CELL 0.05 LBFGS PRINT LSCAL ON CONVERGENCE ADAPT 0.02 CONVERGENCE ENERGY 0.05 PCG TIMESTEP 5 MAXSTEP 10000 MEMORY BIG STEEPEST DESCENT CELL &END &DFT FUNCTIONAL BLYP &END &SYSTEM SYMMETRY 8 CUTOFF 110.0 CELL 10.6 1.0 1.961 0.0 0.0 0.0 &END &ATOMS ... CONSTRAINTS FIX ATOMS 1 1 END CONSTRAINTS &END From sebastia at mpip-mainz.mpg.de Thu Jan 13 15:44:14 2005 From: sebastia at mpip-mainz.mpg.de (Daniel Sebastiani) Date: Thu, 13 Jan 2005 15:44:14 +0100 Subject: [CPMD-list] NMR: 2 possible bugs In-Reply-To: <003701c4eda4$58df86c0$aa00a8c0@UGent.be> References: <003701c4eda4$58df86c0$aa00a8c0@UGent.be> Message-ID: <41E6893E.1050805@mpip-mainz.mpg.de> Dear Reinout, both of your observations are right, there is one bug and one memory overallocation in the NMR current density calculation when combined with the "full" option. This combination has never been used yet (as far as I know). It seems you really digged deep into the response code.... I would be glad to know if you find more bugs :-) >of the code, only 9 x 2 x ngw x nstate is needed with Tcurrent and TNMR_FULL switched > absolutely. Only 9 wfn-arrays are needed at maximum. >H^(p) perturbing Hamiltonian, that is) are stored in column 4-5-6, and then added up to the non-full calculation stored in column 7-8-9. When the subroutine calc_current is then called, the current is calculated from the first six columns, thus using the > This second point has already been corrected - very recently, actually. It will be fixed in the next release of the code. The additional |psi^delta> - orbitals are accumulated in the (7,8,9) rows of C1, together with the |psi^(L)> response orbitals (they have to be processed identically for the current density). Without the FULL option, the |psi^(L)> are stored in rows (4,5,6). This is a little bit confusing, but has "historical" reasons. The simplest bugfix is to define a conditional offset in the current density routine: nmr_current_p: integer i_offs_for_psiL i_offs_for_psiL = 3 if (Tcurrent .AND. Tnmr_full) i_offs_for_psiL = 6 and to modify two lines in nmr_current_p: line 92: [old] call fft2TOr(c0(1,is), psi0, c1(1,is,3+iB), psi1L,psi,ngw,.true.) [new] call fft2TOr(c0(1,is), psi0, [new] c1(1,is,iB+i_offs_for_psiL),psi1L,psi,ngw,.true.) line 96: [old] call apply_op_p(c1(1,is,iB+3), gpsi1L(1,ip),ip, ngw) [new] call apply_op_p(c1(1,is,iB+i_offs_for_psiL),gpsi1L(1,ip),ip, ngw) This should do the job. Happy easter! Daniel -- --------------------------------------------------------- Daniel Sebastiani sebastia at mpip-mainz.mpg.de Max Planck Institute for Polymer Research Ackermannweg 10 Phone +49 6131 379 260 55128 Mainz, Germany Fax +49 6131 379 100 --------------------------------------------------------- From axel.kohlmeyer at theochem.ruhr-uni-bochum.de Thu Jan 13 17:31:05 2005 From: axel.kohlmeyer at theochem.ruhr-uni-bochum.de (Axel Kohlmeyer) Date: Thu, 13 Jan 2005 17:31:05 +0100 Subject: [CPMD-list] error on cell optimization: "need some programming" In-Reply-To: Your message of "Thu, 13 Jan 2005 10:52:55 +0100." <41E644F7.6080407@hmi.de> Message-ID: <200501131631.j0DGV5B06842@yello.theochem.ruhr-uni-bochum.de> >>> "WF" == Wolfgang Fritsch writes: WF> Hi experts, WF> there are already quite a number of entries in the list for the topic of WF> cell optimization. I haven't found anything for my problem, so please WF> bear with me. WF> I am trying to optimize the cell parameters for my case of CuGaSe2. WF> After having, seemingly successfully, optimized the geometry first I WF> then add the keywords WF> CONVERGENCE CELL WF> 0.05 WF> STEEPEST DESCENT CELL WF> and optionally WF> ISOTROPIC CELL WF> But the code stops with error messages after some wavefunction WF> optimization. This is with the mpi version of the code on an alpha cluster. WF> Any suggestion would be appreciated. I append the error message and then WF> the input list. dear wolfgang, a look in the source code (stress.F) will tell you, that the calculation of the stress tensor is not implemented for some conditions: - TLSD (= when using LSDA, i.e. open-shell systems) - TINLC (= when using pseudopotentials with NLCC) - TIVAN (= when using Vanderbilt USPPs). without the stress tensor, you cannot optimize the cell. sorry for the bad news, axel. WF> Thank you for your attention, WF> Wolfgang WF> ================================================================== WF> THIS NEEDS FIRST SOME PROGRAMING ! WF> THIS NEEDS FIRST SOME PROGRAMING ! WF> PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 2] WF> THIS NEEDS FIRST SOME PROGRAMING ! WF> THIS NEEDS FIRST SOME PROGRAMING ! WF> PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 0] WF> THIS NEEDS FIRST SOME PROGRAMING ! WF> PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 4] WF> THIS NEEDS FIRST SOME PROGRAMING ! WF> PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 10] WF> THIS NEEDS FIRST SOME PROGRAMING ! WF> ...and a few more of the kind, differreing by the number after PROC= WF> ====================================================================== WF> &CPMD WF> OPTIMIZE GEOMETRY WF> CONVERGENCE GEOMETRY WF> 0.0005 WF> CONVERGENCE ORBITALS WF> 0.0005 WF> CONVERGENCE CELL WF> 0.05 WF> LBFGS WF> PRINT LSCAL ON WF> CONVERGENCE ADAPT WF> 0.02 WF> CONVERGENCE ENERGY WF> 0.05 WF> PCG WF> TIMESTEP WF> 5 WF> MAXSTEP WF> 10000 WF> MEMORY BIG WF> STEEPEST DESCENT CELL WF> &END WF> &DFT WF> FUNCTIONAL BLYP WF> &END WF> &SYSTEM WF> SYMMETRY WF> 8 WF> CUTOFF WF> 110.0 WF> CELL WF> 10.6 1.0 1.961 0.0 0.0 0.0 WF> &END WF> &ATOMS WF> ... WF> CONSTRAINTS WF> FIX ATOMS WF> 1 WF> 1 WF> END CONSTRAINTS WF> &END WF> _______________________________________________ WF> CPMD-list mailing list WF> CPMD-list at cpmd.org WF> http://cpmd.org/mailman/listinfo/cpmd-list -- ======================================================================= Axel Kohlmeyer e-mail: axel.kohlmeyer at theochem.ruhr-uni-bochum.de Lehrstuhl fuer Theoretische Chemie Phone: ++49 (0)234/32-26673 Ruhr-Universitaet Bochum - NC 03/53 Fax: ++49 (0)234/32-14045 D-44780 Bochum http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/ ======================================================================= If you make something idiot-proof, the universe creates a better idiot. From hutter at pci.unizh.ch Thu Jan 13 18:00:45 2005 From: hutter at pci.unizh.ch (Juerg Hutter) Date: Thu, 13 Jan 2005 18:00:45 +0100 (MET) Subject: [CPMD-list] Pseudopotential of Chlorine In-Reply-To: <1105461339.6339.12.camel@phpc32.ph.kcl.ac.uk> References: <1105461339.6339.12.camel@phpc32.ph.kcl.ac.uk> Message-ID: Hi attached is a PP generated with the same cutoff values as the one in the library for revPBE (Cl_MT_revPBE.psp). No warrenty that this is any good. TEST IT YOURSELF! regards Juerg Hutter PS: you can find the Hamann atomic code and a PP library at my web page http://www.pci.unizh.ch/gruppe.hutter/e/information.html ---------------------------------------------------------- Juerg Hutter Phone : ++41 1 635 4491 Physical Chemistry Institute FAX : ++41 1 635 6838 University of Zurich E-mail: hutter at pci.unizh.ch Winterthurerstrasse 190 CH-8057 Zurich, Switzerland ---------------------------------------------------------- On Tue, 11 Jan 2005, Claudio Melis wrote: > Dear CPMD users, > I'm looking for a Martin Troullier pseudopotential with the PBE > functional for chlorine. > Does anybody know where I can find it ? > Thanks > Claudio Melis > > _______________________________________________ > CPMD-list mailing list > CPMD-list at cpmd.org > http://cpmd.org/mailman/listinfo/cpmd-list > -------------- next part -------------- A non-text attachment was scrubbed... Name: Cl_MT_PBE.gz Type: application/octet-stream Size: 35202 bytes Desc: Url : http://cpmd.org/pipermail/cpmd-list/attachments/20050113/182da8b0/attachment.obj From hutter at pci.unizh.ch Thu Jan 13 18:02:58 2005 From: hutter at pci.unizh.ch (Juerg Hutter) Date: Thu, 13 Jan 2005 18:02:58 +0100 (MET) Subject: [CPMD-list] error on cell optimization: "need some programming" In-Reply-To: <41E644F7.6080407@hmi.de> References: <41E644F7.6080407@hmi.de> Message-ID: Hi because of lack of support the stress calculation for the options LSD NLCC USPP has been disabled until somebody will (re)debug the code. regards Juerg Hutter ---------------------------------------------------------- Juerg Hutter Phone : ++41 1 635 4491 Physical Chemistry Institute FAX : ++41 1 635 6838 University of Zurich E-mail: hutter at pci.unizh.ch Winterthurerstrasse 190 CH-8057 Zurich, Switzerland ---------------------------------------------------------- On Thu, 13 Jan 2005, Wolfgang Fritsch wrote: > Hi experts, > > there are already quite a number of entries in the list for the topic of > cell optimization. I haven't found anything for my problem, so please > bear with me. > > I am trying to optimize the cell parameters for my case of CuGaSe2. > After having, seemingly successfully, optimized the geometry first I > then add the keywords > > CONVERGENCE CELL > 0.05 > > STEEPEST DESCENT CELL > > and optionally > > ISOTROPIC CELL > > But the code stops with error messages after some wavefunction > optimization. This is with the mpi version of the code on an alpha cluster. > > Any suggestion would be appreciated. I append the error message and then > the input list. > > Thank you for your attention, > > Wolfgang > > ================================================================== > THIS NEEDS FIRST SOME PROGRAMING ! > THIS NEEDS FIRST SOME PROGRAMING ! > > > PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 2] > THIS NEEDS FIRST SOME PROGRAMING ! > THIS NEEDS FIRST SOME PROGRAMING ! > > > PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 0] > THIS NEEDS FIRST SOME PROGRAMING ! > > > PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 4] > THIS NEEDS FIRST SOME PROGRAMING ! > > > PROGRAM STOPS IN SUBROUTINE STRESS| [PROC= 10] > THIS NEEDS FIRST SOME PROGRAMING ! > > ...and a few more of the kind, differreing by the number after PROC= > > > ====================================================================== > > &CPMD > OPTIMIZE GEOMETRY > CONVERGENCE GEOMETRY > 0.0005 > CONVERGENCE ORBITALS > 0.0005 > CONVERGENCE CELL > 0.05 > LBFGS > PRINT LSCAL ON > CONVERGENCE ADAPT > 0.02 > CONVERGENCE ENERGY > 0.05 > PCG > TIMESTEP > 5 > MAXSTEP > 10000 > MEMORY BIG > STEEPEST DESCENT CELL > &END > &DFT > FUNCTIONAL BLYP > &END > &SYSTEM > SYMMETRY > 8 > CUTOFF > 110.0 > CELL > 10.6 1.0 1.961 0.0 0.0 0.0 > &END > > &ATOMS > ... > > CONSTRAINTS > FIX ATOMS > 1 > 1 > END CONSTRAINTS > > &END > > _______________________________________________ > CPMD-list mailing list > CPMD-list at cpmd.org > http://cpmd.org/mailman/listinfo/cpmd-list > From ZRLSRB at ch.ibm.com Fri Jan 14 09:50:46 2005 From: ZRLSRB at ch.ibm.com (Salomon Billeter) Date: Fri, 14 Jan 2005 09:50:46 +0100 Subject: Bug in pcgrad.F [Re: [CPMD-list] slow geometry optimization - would underrelaxation help?] Message-ID: Dear Wolfgang, the bug causing the program to stop is in the line search routine of the preconditioned-conjugate-gradient wavefunction optimization (pcgrad.F) which has been improved lately. Since the total energy (with the forces) may be calculated there, it is no longer sufficient to back up the energies only on one processor. You can just delete the IF (PARENT) condition on the calls to EBACK. Below is a patch file. I send this message also to the list. Thanks again for sending the inputs. Best regards, Salomon (See attached file: pcgrad.patch) -------------- next part -------------- A non-text attachment was scrubbed... Name: pcgrad.patch Type: application/octet-stream Size: 1443 bytes Desc: not available Url : http://cpmd.org/pipermail/cpmd-list/attachments/20050114/3e8d5411/attachment.obj From bob at bob.usuhs.mil Fri Jan 14 15:28:05 2005 From: bob at bob.usuhs.mil (Robert Williams) Date: Fri, 14 Jan 2005 09:28:05 -0500 Subject: [CPMD-list] Problems optimizing monoclinic unit cell Message-ID: <41E7D6F5.9030900@bob.usuhs.mil> Dear List, I encountered the same problem as posted earlier by Ewald Pauwels (see below) and was initially confused because the expected form in CPMD: a b c d 0.0 0.0 doesn't follow convention, as in http://xrayweb.msg.ku.edu/notes/symmetry.html : "Certain simple conventions have been followed in tabulating the parameters. In the monoclinic system, one of the axes is unique in the sense that it is perpendicular to the other two axes. This axis is selected by convention as either b or c so that either ? or ? are ? 90?, respectively. Note that c unique monoclinic cells are common in French literature and b unique cells are common in most other languages... Triclinic 6 a ? b ? c; ? ? ? ? ? Monoclinic 4 a ? b ? c; ? = ? = 90?, ? ? 90? Orthorhombic 3 a ? b ? c; ? = ? = ? = 90? ..." etc. I'm assuming that CPMD expects "a" unique cells (in the interest of international peace) so I've started a test job where my original "b" unique cell has a and b switched, and ? and ? switched, (I hope my character encoding for alpha "?" and beta "?" work for you) and where the Cartesian coordinates have columns x and y switched. At least this job succeeded past the error message with respect to symmetry, but it failed with a memory allocation error which I will post in a separate message. Please correct me if I am wrong in my assumptions or in my simple minded solution. If this is correct, it would be useful to have an explanation in the manual about what must be done to change c and d unique cells to the expected form. Bob Williams Ewald Pauwels wrote: >I first started out with the definition for my molecular system using >the keywords SYMMETRY and CELL: > &SYSTEM > ANGSTROM > SYMMETRY > 12 > CELL > 5.0872 2.314338 1.073321 0.0 -0.374493314 0.0 > >As Bala Sundaram pointed out to me, this specific format (as given in >the manual) was incorrect and should be replaced by: > &SYSTEM > ANGSTROM > SYMMETRY > 12 > CELL > 5.0872 2.314338 1.073321 -0.374493314 0.0 0.0 >That is, the format of the CELL keyword for a monoclinic system is: > a b/a c/a cos(beta) 0 0 >A notification of this format is also implemented in version 3.9.1 of >cpmd, although I think some mention in the manual would be helpfull. >Anyhow, the calculation still failed..... -- Dr. Robert Williams Dept. of Biomedical Informatics Uniformed Services University 301-295-3568 From axel.kohlmeyer at theochem.ruhr-uni-bochum.de Fri Jan 14 16:49:04 2005 From: axel.kohlmeyer at theochem.ruhr-uni-bochum.de (Axel Kohlmeyer) Date: Fri, 14 Jan 2005 16:49:04 +0100 Subject: [CPMD-list] Problems optimizing monoclinic unit cell In-Reply-To: Your message of "Fri, 14 Jan 2005 09:28:05 EST." <41E7D6F5.9030900@bob.usuhs.mil> Message-ID: <200501141549.j0EFn4U08056@yello.theochem.ruhr-uni-bochum.de> >>> "RW" == Robert Williams writes: RW> Dear List, RW> I encountered the same problem as posted earlier RW> by Ewald Pauwels (see below) and was initially RW> confused because the expected form in CPMD: a b c d 0.0 0.0 dear bob, as you have already found out, there are so many conventions to describe a unit cell, that there is always the chance, that you mix something up when moving data. my knowledge about crystallography is limited, but here is a summary that works for me: CELL expects: a b/a c/a cos(alpha) cos(beta) cos(gamma) CELL ABSOLUTE expects: a b c cos(alpha) cos(beta) cos(gamma) CELL DEGREE expects: a b/a c/a alpha beta gamma CELL ABSOLUTE DEGREE expects: a b c alpha beta gamma when you run out of memory, this frequently happens when you give 'a b c' but do not use CELL ABSOLUTE. the next thing, and this is the one that irritated me the most, is the fact, that you always have to give the cell dimensions for the corresponding _conventional_ cell, but the coordinates for the _primitive_ cell. see for example the inputs and outputs at (which on top of that use fractional coordinates with SCALE): http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/cpmd-tutor/reference/05-si_bulk/ if you compute the total energy per atom, you'll see that the values agree (provided the k-point sampling is good enough). most of that information should already be somewhere in the manual at some place or another. feel free to provide an improved wording or contribute some additional text. in my experience, the best way to deal with this issue is to provide some examples the work (see above). regards, axel. RW> doesn't follow convention, as in RW> http://xrayweb.msg.ku.edu/notes/symmetry.html : RW> "Certain simple conventions have been followed in tabulating RW> the parameters. In the monoclinic system, one of the axes is RW> unique in the sense that it is perpendicular to the other two RW> axes. This axis is selected by convention as either b or c so RW> that either ?? or ?? are ??? 90??, respectively. Note that c unique RW> monoclinic cells are common in French literature and b unique RW> cells are common in most other languages... RW> Triclinic 6 a ??? b ??? c; ?? ??? ?? ??? ?? RW> Monoclinic 4 a ??? b ??? c; ?? = ?? = 90??, ?? ??? 90?? RW> Orthorhombic 3 a ??? b ??? c; ?? = ?? = ?? = 90?? RW> ..." etc. RW> I'm assuming that CPMD expects "a" unique cells RW> (in the interest of international peace) so RW> I've started a test job where my original "b" unique cell RW> has a and b switched, and ?? and ?? switched, RW> (I hope my character encoding for alpha "??" and beta "??" work for you) RW> and where the Cartesian coordinates have columns x and y switched. RW> At least this job succeeded past the error message with respect to RW> symmetry, but it failed with a memory allocation error RW> which I will post in a separate message. Please RW> correct me if I am wrong in my assumptions or in RW> my simple minded solution. RW> If this is correct, it would be useful to have an RW> explanation in the manual about what must be done RW> to change c and d unique cells to the expected form. RW> Bob Williams [...] -- ======================================================================= Axel Kohlmeyer e-mail: axel.kohlmeyer at theochem.ruhr-uni-bochum.de Lehrstuhl fuer Theoretische Chemie Phone: ++49 (0)234/32-26673 Ruhr-Universitaet Bochum - NC 03/53 Fax: ++49 (0)234/32-14045 D-44780 Bochum http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/ ======================================================================= If you make something idiot-proof, the universe creates a better idiot. From bob at bob.usuhs.mil Fri Jan 14 17:16:40 2005 From: bob at bob.usuhs.mil (Robert Williams) Date: Fri, 14 Jan 2005 11:16:40 -0500 Subject: [CPMD-list] ALLOCATION OF MEMORY FAILED Message-ID: <41E7F068.5010307@bob.usuhs.mil> Dear list, This problem on 32 bit intel is biting me again, so I've put together some information that may help others. I, however, still need some help with this, since I think I have access to more memory than CPMD is able to allocate. I can allocate at least 2.5 Gbytes with a simple ifc fortran program (see below), but (I think) CPMD is failing to allocate memory that is below 1.7 Gbytes. Please check my logs below and correct me if I'm wrong. ------------------------------------------------ This issue with 32 bit processors is cropping up more frequently recently in scientific computing. Documentation about it on the web has been sparse, so I've gathered a few links that address this problem. The best is http://www.spack.org/wiki/LinuxRamLimits where it is made more or less clear that there is a hard and fast limit of around 3 Gbytes or a little less for a single process on 32 bit intel. Also good, including a simple fortran program to test memory access: http://softwareforums.intel.com/ids/board/message?board.id=11&message.id=100 -------------------------------------------------------------- So...my problem, if I can access 2.5 Gbytes with other programs, why this on a serial cpmd compiled with ifort 8.1: *** RGGEN| THE NEW SIZE OF THE PROGRAM IS 92896/ 111440 kBYTES *** *** RINFORCE| THE NEW SIZE OF THE PROGRAM IS 271112/ 290060 kBYTES *** *** FFTPRP| THE NEW SIZE OF THE PROGRAM IS 482992/ 500276 kBYTES *** ... **************************************************************** PROCESSOR 0 ALLOCATION OF 146341736 WORDS OF MEMORY FAILED **************************************************************** *** MEMORY| THE NEW SIZE OF THE PROGRAM IS 506240/1669568 kBYTES *** ... [PEAK NUMBER 54] PEAK MEMORY 210876035 = 1687.0 MBytes top reports: Mem: 3628760k total, 3116784k used, 511976k free, 2668k buffers Swap: 5040228k total, 4k used, 5040224k free, 2945044k cached [root at linux1 linux-2.4.21]# ulimit -a core file size (blocks, -c) 0 data seg size (kbytes, -d) unlimited file size (blocks, -f) unlimited max locked memory (kbytes, -l) unlimited max memory size (kbytes, -m) unlimited open files (-n) 1024 pipe size (512 bytes, -p) 8 stack size (kbytes, -s) 327680 cpu time (seconds, -t) unlimited max user processes (-u) 28657 virtual memory (kbytes, -v) unlimited The memtest program to test how much memory can be allocated: [bob at linux1 aaa_ap]$ ./memtest A( 1 ): ok - memory allocated (GB) = 0.2500000 B( 1 ): ok - memory allocated (GB) = 0.5000000 C( 1 ): ok - memory allocated (GB) = 0.7500000 A( 2 ): ok - memory allocated (GB) = 1.000000 B( 2 ): ok - memory allocated (GB) = 1.250000 C( 2 ): ok - memory allocated (GB) = 1.500000 A( 3 ): ok - memory allocated (GB) = 1.750000 B( 3 ): ok - memory allocated (GB) = 2.000000 C( 3 ): ok - memory allocated (GB) = 2.250000 A( 4 ): ok - memory allocated (GB) = 2.500000 Allocation of B( 4 ) failed with status= 41 Thanks in -- Dr. Robert Williams Dept. of Biomedical Informatics Uniformed Services University 301-295-3568 From axel.kohlmeyer at theochem.ruhr-uni-bochum.de Fri Jan 14 18:58:43 2005 From: axel.kohlmeyer at theochem.ruhr-uni-bochum.de (Axel Kohlmeyer) Date: Fri, 14 Jan 2005 18:58:43 +0100 Subject: [CPMD-list] ALLOCATION OF MEMORY FAILED In-Reply-To: Your message of "Fri, 14 Jan 2005 11:16:40 EST." <41E7F068.5010307@bob.usuhs.mil> Message-ID: <200501141758.j0EHwhW14713@yello.theochem.ruhr-uni-bochum.de> >>> "RW" == Robert Williams writes: RW> Dear list, dear bob, RW> This problem on 32 bit intel is biting me again, RW> so I've put together some information that may help others. RW> I, however, still need some help with this, RW> since I think I have access to more memory than RW> CPMD is able to allocate. RW> I can allocate at least 2.5 Gbytes with a simple RW> ifc fortran program (see below), RW> but (I think) CPMD is failing RW> to allocate memory that is below 1.7 Gbytes. RW> Please check my logs below and correct me if I'm wrong. according to the output below, you want to allocate 146341736 real*8 words = 1.1GB on top of the already allocated ~1.7GB. so you _are_ above the 2GB mark. RW> ------------------------------------------------ RW> This issue with 32 bit processors is cropping up RW> more frequently recently in scientific computing. RW> Documentation about it on the web has been sparse, RW> so I've gathered a few links that address this RW> problem. The best is RW> http://www.spack.org/wiki/LinuxRamLimits RW> where it is made more or less clear that there RW> is a hard and fast limit of around 3 Gbytes RW> or a little less for RW> a single process on 32 bit intel. this is the maximum available address space for a single process using a 'normal' kernel with ~1GB physical memory. this can be changed to a 2:2 setting or even a 1:3 setting (1 GB address space per process and 3GB physical). with the 'bigmem' setup you use instructions in newer cpus that allow access to memory pages beyond that, but still the limits from above have to be fulfilled, so the kernel can only give access to some of these extended pages at the same time. RW> Also good, including a simple fortran program to RW> test memory access: RW> http://softwareforums.intel.com/ids/board/message?board.id=11&message.id=100 RW> -------------------------------------------------------------- note, that this code is written in fortran 90 and that it uses the fortran 90 memory allocation procedure. RW> So...my problem, if I can access 2.5 Gbytes with other programs, RW> why this on a serial cpmd compiled with ifort 8.1: RW> *** RGGEN| THE NEW SIZE OF THE PROGRAM IS 92896/ 111440 kBYTES *** RW> *** RINFORCE| THE NEW SIZE OF THE PROGRAM IS 271112/ 290060 kBYTES *** RW> *** FFTPRP| THE NEW SIZE OF THE PROGRAM IS 482992/ 500276 kBYTES *** RW> ... RW> **************************************************************** RW> PROCESSOR 0 ALLOCATION OF 146341736 WORDS OF MEMORY FAILED RW> **************************************************************** RW> *** MEMORY| THE NEW SIZE OF THE PROGRAM IS 506240/1669568 kBYTES *** RW> ... RW> [PEAK NUMBER 54] PEAK MEMORY 210876035 = 1687.0 MBytes most likely because CPMD is using 'Cray Pointers' to store memory adresses. the address returned is stored in an integer and since integers in fortran are signed, every address above 2GB will be negative. also the fortran runtime has to be completely free of any signed vs. unsigned issues (which seems to be the case, since you _can_ get 2.5GB with the f90 test program). btw, this had happened some time ago with the pgf90 (but not pgf77 compiler). have a look at the code in memory.F around line 100 you have: IF(IP_XM.LE.0) THEN IERROR=1 ENDIF you can try to eliminate this test (or change the test to IP_XM.EQ 0) to get 1GB more. you may get lucky. better try to run this job on a 64-bit machine or run two MPI nodes per processor. regards, axel. [...] RW> Thanks in RW> -- RW> Dr. Robert Williams RW> Dept. of Biomedical Informatics RW> Uniformed Services University RW> 301-295-3568 RW> _______________________________________________ RW> CPMD-list mailing list RW> CPMD-list at cpmd.org RW> http://cpmd.org/mailman/listinfo/cpmd-list -- ======================================================================= Axel Kohlmeyer e-mail: axel.kohlmeyer at theochem.ruhr-uni-bochum.de Lehrstuhl fuer Theoretische Chemie Phone: ++49 (0)234/32-26673 Ruhr-Universitaet Bochum - NC 03/53 Fax: ++49 (0)234/32-14045 D-44780 Bochum http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/ ======================================================================= If you make something idiot-proof, the universe creates a better idiot. From wolfi at mittelerde.physik.uni-konstanz.de Mon Jan 17 20:43:05 2005 From: wolfi at mittelerde.physik.uni-konstanz.de (Wolfram Quester) Date: Mon, 17 Jan 2005 20:43:05 +0100 Subject: [CPMD-list] Problems optimizing geometry using FIX STRUCTURE DIST Message-ID: <20050117194305.GA20539@halley.zuhause> Dear CPMD users and developers, I try to calculate bonding energies of Si-clusters. To give predefined distances between the clusters I have in my input file &SYSTEM ANGSTROM SYMMETRY 0 POISSON SOLVER TUCKERMAN CELL ABSOLUTE DEGREE 40.0 20.0 20.0 90 90 90 CUTOFF 20.0 &END &ATOMS Silicon [...snip...] CONSTRAINTS FIX STRUCTURE 1 DIST 1 8 1.200000 END CONSTRAINTS &END From the documentation I expected that these constraints keep the distance between atoms 1 and 8 fixed during the simulation run. But in my output I have in the beginning <<<<< CONSTRAINTS >>>>> FIXED STRUCTURE ELEMENTS TYPE ATOM ATOM ATOM ATOM VALUE DIFFERENCE DISTANCE 1 8 0 0 2.26767 -1.776357E-15 and in the end: <<<<< CONSTRAINTS >>>>> FIXED STRUCTURE ELEMENTS TYPE ATOM ATOM ATOM ATOM VALUE DIFFERENCE DISTANCE 1 8 0 0 2.92527 6.575977E-01 This means that that now the distance between the two atoms is 0.658 Bohr larger than in the beginning. But if it is "fixed", this value shouldn't change? Or am I wrong? And I don't get a file CONSTRAINT created, which should be there. Thanks for your hints, Wolfi -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: Digital signature Url : http://cpmd.org/pipermail/cpmd-list/attachments/20050117/ab50fa49/attachment.pgp From Ari.P.Seitsonen at iki.fi Mon Jan 17 22:48:09 2005 From: Ari.P.Seitsonen at iki.fi (Ari P Seitsonen) Date: Mon, 17 Jan 2005 22:48:09 +0100 (CET) Subject: [CPMD-list] Problems optimizing geometry using FIX STRUCTURE DIST In-Reply-To: <20050117194305.GA20539@halley.zuhause> References: <20050117194305.GA20539@halley.zuhause> Message-ID: Dear Wolfram, CPMD has sometimes problems with geometry optimisation and constraints; it's more typical to use molecular dynamics and usually simulated annealing, employing very low temperatures (and finally reaching practically zero). Some one who knows this part better can probably give a better explanation. Greetings from Zurich, apsi PS It's usually advantageous to use the 'LBFGS' for geometry optimisations, in case you didn't specify it already -=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=- Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/ Tel +41 1 635 44 97 / Fax +41 1 635 68 38 / GSM +41 79 719 09 35 Anschrift: Physikalisch Chemisches Institut (PCI), Universit?t Z?rich (UniZh) Indirizzo: Winterthurerstrasse 190, CH-8057 Z?rich Address: Schweiz / Svizzera / Suisse / Svizra / Switzerland On Mon, 17 Jan 2005, Wolfram Quester wrote: > Dear CPMD users and developers, > > I try to calculate bonding energies of Si-clusters. To give predefined > distances between the clusters I have in my input file > > &SYSTEM > ANGSTROM > SYMMETRY > 0 > POISSON SOLVER TUCKERMAN > CELL ABSOLUTE DEGREE > 40.0 20.0 20.0 90 90 90 > CUTOFF > 20.0 > &END > > &ATOMS > Silicon > [...snip...] > CONSTRAINTS > FIX STRUCTURE > 1 > DIST 1 8 1.200000 > END CONSTRAINTS > &END > > From the documentation I expected that these constraints keep the > distance between atoms 1 and 8 fixed during the simulation run. But in > my output I have in the beginning > > > <<<<< CONSTRAINTS >>>>> > FIXED STRUCTURE ELEMENTS > TYPE ATOM ATOM ATOM ATOM VALUE DIFFERENCE > DISTANCE 1 8 0 0 2.26767 -1.776357E-15 > > and in the end: > > <<<<< CONSTRAINTS >>>>> > FIXED STRUCTURE ELEMENTS > TYPE ATOM ATOM ATOM ATOM VALUE DIFFERENCE > DISTANCE 1 8 0 0 2.92527 6.575977E-01 > > This means that that now the distance between the two atoms is 0.658 > Bohr larger than in the beginning. But if it is "fixed", this value > shouldn't change? Or am I wrong? And I don't get a file CONSTRAINT > created, which should be there. > > Thanks for your hints, > > Wolfi > > From audrius.alkauskas at unibas.ch Tue Jan 18 13:38:40 2005 From: audrius.alkauskas at unibas.ch (Audrius Alkauskas) Date: Tue, 18 Jan 2005 13:38:40 +0100 Subject: [CPMD-list] Surface in the 3.9 version Message-ID: <41ED0350.1090304@unibas.ch> Hello I just wonder whether SURFACE option works with non-orthorhombic unit cells (for instance, hexagonal) in the new (3.9) version of CPMD. Thank you Audrius From ZRLSRB at ch.ibm.com Tue Jan 18 14:03:04 2005 From: ZRLSRB at ch.ibm.com (Salomon Billeter) Date: Tue, 18 Jan 2005 14:03:04 +0100 Subject: [CPMD-list] Surface in the 3.9 version Message-ID: Dear Audrius, hexagonal cells work with SURFACE starting from CPMD version 3.9.1. Older versions had a bug preventing this. Starting from version 3.9.1, you can also use HOCKNEY with SURFACE to save memory and CPU time (it's not a real Hockney Poisson solver, since the influence function is prepared in reciprocal space, but you get away with the box requirement of Hockney, i.e. the density must be zero at the border). In the tests, this functionality was working correctly, but please double-check your results. Best regards, Salomon From binpan at MIT.EDU Thu Jan 20 05:36:44 2005 From: binpan at MIT.EDU (Bin Pan) Date: Wed, 19 Jan 2005 23:36:44 -0500 Subject: [CPMD-list] A problem with memory allocation Message-ID: <5.2.1.1.2.20050119233018.01f37ec0@po10.mit.edu> Dear all, I just compiled a parallel version of cpmd3.9 using ifc and mpich. But after I executed the parallel command, I found the output is like the following. Can anyone please tell me how to solve this problem? BTW, I suspect that is relevant to the fact that the system is 64bit. But I really do not know exactly what caused this. The installation process for ifc 8.0 using 32bit was successful on this machine. I used the non-commercial version of this fortran compiler from their website. Thank you very much!!! ================================================================ = UNKNOWN KEYWORDS IN SECTION &CPMD = = mOLECULAR DYNAMICS = ================================================================ EXCHANGE CORRELATION FUNCTIONALS LDA EXCHANGE: SLATER (ALPHA = 0.66667) LDA CORRELATION: LEE, YANG & PARR [C.L. LEE, W. YANG, AND R.G. PARR, PRB 37 785 (1988)] GRADIENT CORRECTED FUNCTIONAL DENSITY THRESHOLD: 1.00000E-08 EXCHANGE ENERGY [A.D. BECKE, PHYS. REV. A 38, 3098 (1988)] PARAMETER BETA: 0.004200 CORRELATION ENERGY [LYP: C.L. LEE ET AL. PHYS. REV. B 37, 785 (1988)] **************************************************************** PROCESSOR 0 ALLOCATION OF 321280 WORDS OF MEMORY FAILED **************************************************************** *** MEMORY| THE NEW SIZE OF THE PROGRAM IS 7232/ 19284 kBYTES *** ================================================================ BIG MEMORY ALLOCATIONS NGHTOL 400 WGH 400 RGH 400 WSG 400 NGHCOM 400 VELP 108 TAU0 108 RCL 48 BL 48 AL 48 ---------------------------------------------------------------- [PEAK NUMBER 11] PEAK MEMORY 2400 = 0.0 MBytes ================================================================ PROGRAM STOPS IN SUBROUTINE MEMORY| ALLOCATION FAILED (GNL) [PROC= 0] p0_16389: p4_error: : 999 From binpan at MIT.EDU Thu Jan 20 19:36:35 2005 From: binpan at MIT.EDU (Bin Pan) Date: Thu, 20 Jan 2005 13:36:35 -0500 Subject: [CPMD-list] A problem with memory allocation In-Reply-To: <20050120070025.17161.qmail@web25710.mail.ukl.yahoo.com> References: <5.2.1.1.2.20050119233018.01f37ec0@po10.mit.edu> Message-ID: <5.2.1.1.2.20050120133443.01f50110@po10.mit.edu> Thank you Abraham! I thought this problem comes from the intel fortran compiler, because it is Intel(R) Fortran Compiler for 32-bit applications, Version 8.1 which is not suitable for 64bit machines. I tried almost all the ways mentioned on the forum, just did not work. Any further suggestion will be greatly appreciated! Best, Bin. At 07:00 AM 1/20/2005 +0000, you wrote: >You could try a google search of the forum. It might >be a problem related to the way ifc 8.0 handles the >threads lib (libpthreads) in linux. > >Alternatively you can try setting(i.e execute in >terminal): > >ulimit -s unlimited > >This is more likely to work for the serial rather than >the parallel implementation. > > --- Bin Pan wrote: > > Dear all, > > > > I just compiled a parallel version of cpmd3.9 using > > ifc and mpich. > > But after I executed the parallel command, I found > > the output is like the > > following. > > Can anyone please tell me how to solve this problem? > > BTW, I suspect that is relevant to the fact that the > > system is 64bit. But I > > really do not know > > exactly what caused this. The installation process > > for ifc 8.0 using 32bit > > was successful on this machine. > > I used the non-commercial version of this fortran > > compiler from their website. > > > > Thank you very much!!! > > > > >================================================================ > > = UNKNOWN KEYWORDS IN SECTION &CPMD > > = > > = mOLECULAR DYNAMICS > > = > > > > >================================================================ > > > > > > EXCHANGE CORRELATION FUNCTIONALS > > LDA EXCHANGE: SLATER > > (ALPHA = 0.66667) > > LDA CORRELATION: > > LEE, YANG & PARR > > [C.L. LEE, W. YANG, AND R.G. PARR, PRB 37 > > 785 (1988)] > > GRADIENT CORRECTED FUNCTIONAL > > DENSITY THRESHOLD: > > 1.00000E-08 > > EXCHANGE ENERGY > > [A.D. BECKE, PHYS. REV. A 38, 3098 (1988)] > > PARAMETER BETA: > > 0.004200 > > CORRELATION ENERGY > > [LYP: C.L. LEE ET AL. PHYS. REV. B 37, 785 > > (1988)] > > > > > > > > >**************************************************************** > > PROCESSOR 0 ALLOCATION OF 321280 WORDS OF > > MEMORY FAILED > > > > >**************************************************************** > > > > *** MEMORY| THE NEW SIZE OF THE PROGRAM IS > > 7232/ 19284 kBYTES *** > > > > > > >================================================================ > > BIG MEMORY ALLOCATIONS > > NGHTOL 400 WGH > > 400 > > RGH 400 WSG > > 400 > > NGHCOM 400 VELP > > 108 > > TAU0 108 RCL > > 48 > > BL 48 AL > > 48 > > > > >---------------------------------------------------------------- > > [PEAK NUMBER 11] PEAK MEMORY 2400 = > > 0.0 MBytes > > > > >================================================================ > > > > > > PROGRAM STOPS IN SUBROUTINE MEMORY| ALLOCATION > > FAILED (GNL) [PROC= 0] > > p0_16389: p4_error: : 999 > > > > _______________________________________________ > > CPMD-list mailing list > > CPMD-list at cpmd.org > > http://cpmd.org/mailman/listinfo/cpmd-list > > > > > > > >___________________________________________________________ >ALL-NEW Yahoo! Messenger - all new features - even more fun! >http://uk.messenger.yahoo.com From aklix at utnl.jp Fri Jan 21 01:51:16 2005 From: aklix at utnl.jp (Axel Klix) Date: Fri, 21 Jan 2005 09:51:16 +0900 Subject: [CPMD-list] Re: CPMD-list Digest, Vol 9, Issue 9 In-Reply-To: <200501201100.MAA16926@cpmd.org> References: <200501201100.MAA16926@cpmd.org> Message-ID: <41F05204.7040409@utnl.jp> I think this is a problem with the MPI and you need to change/set the P4_GLOBMEMSIZE variable in your shell to a high enough value. Cheers! Axel > > ------------------------------------------------------------------------ > > Subject: > [CPMD-list] A problem with memory allocation > From: > Bin Pan > Date: > Wed, 19 Jan 2005 23:36:44 -0500 > To: > cpmd-list at cpmd.org > > To: > cpmd-list at cpmd.org > > > Dear all, > > I just compiled a parallel version of cpmd3.9 using ifc and mpich. > But after I executed the parallel command, I found the output is like > the following. > Can anyone please tell me how to solve this problem? > BTW, I suspect that is relevant to the fact that the system is 64bit. > But I really do not know > exactly what caused this. The installation process for ifc 8.0 using > 32bit was successful on this machine. > I used the non-commercial version of this fortran compiler from their > website. > > Thank you very much!!! > > ================================================================ > = UNKNOWN KEYWORDS IN SECTION &CPMD = > = mOLECULAR DYNAMICS = > ================================================================ > > > EXCHANGE CORRELATION FUNCTIONALS > LDA EXCHANGE: SLATER (ALPHA = 0.66667) > LDA CORRELATION: LEE, YANG & PARR > [C.L. LEE, W. YANG, AND R.G. PARR, PRB 37 785 (1988)] > GRADIENT CORRECTED FUNCTIONAL > DENSITY THRESHOLD: 1.00000E-08 > EXCHANGE ENERGY > [A.D. BECKE, PHYS. REV. A 38, 3098 (1988)] > PARAMETER BETA: 0.004200 > CORRELATION ENERGY > [LYP: C.L. LEE ET AL. PHYS. REV. B 37, 785 (1988)] > > > **************************************************************** > PROCESSOR 0 ALLOCATION OF 321280 WORDS OF MEMORY FAILED > **************************************************************** > > *** MEMORY| THE NEW SIZE OF THE PROGRAM IS 7232/ 19284 kBYTES *** > > ================================================================ > BIG MEMORY ALLOCATIONS > NGHTOL 400 WGH 400 > RGH 400 WSG 400 > NGHCOM 400 VELP 108 > TAU0 108 RCL 48 > BL 48 AL 48 > ---------------------------------------------------------------- > [PEAK NUMBER 11] PEAK MEMORY 2400 = 0.0 MBytes > ================================================================ > > > PROGRAM STOPS IN SUBROUTINE MEMORY| ALLOCATION FAILED (GNL) [PROC= 0] > p0_16389: p4_error: : 999 > > >------------------------------------------------------------------------ > >_______________________________________________ >CPMD-list mailing list >CPMD-list at cpmd.org >http://cpmd.org/mailman/listinfo/cpmd-list > > -- Axel Klix University of Tokyo, NERL Tokaimura Phone: +81 - (0)29-287-8918 email: aklix at utnl.jp From shyma at imr.ac.cn Tue Jan 25 02:17:13 2005 From: shyma at imr.ac.cn (=?gb2312?B?c2h5bWE=?=) Date: Tue, 25 Jan 2005 09:17:13 +0800 (CST) Subject: [CPMD-list] Compiling errors, help! Message-ID: <41F59E19.0000BA.23879@app-01> Dear Sir: I'm a newer on CPMD. The version of Linux I installed was Redhat 9.0 . I have compiled a serial CPMD3.9 (the platform I chosen was PC-PGI) executable for an Intel4 processor with the pgf90, but when I tried to compiled a parallel CPMD3.9 with pgf90 (I chosen the PC-PGI-MPI platform), the error such as ?? rm -f cpmd.f /lib/cpp -P -C -traditional -D__Linux -D__PGI -DLAPACK -DFFT_DEFAULT -DPARALLEL -DMP_LIBRARY=__MPI -DMYRINET /home/shyma/cpmd/P-CPMD-3.9.1/SOURCE/cpmd.F ./cpmd.f mpif77 -c -fast -I. -I/home/shyma/cpmd/P-CPMD-3.9.1/SOURCE -Mr8 -pc64 ./cpmd.f -o ./cpmd.o g77: unrecognized option `-Mr8' g77: unrecognized option `-pc64' f771: unrecognized option `-fast' mpif77: No such file or directory make: *** [cpmd.o] Error 1 ?? displayed. TO find a solution, I change the ??Makefile??(in fact i didn't understand it clearly) as fellows "???? # SHELL = /bin/sh # #--------------- Default Configuration for pc-pgi-mpi --------------- SRC = . DEST = . BIN = . #QMMM_FLAGS = -D__QMECHCOUPL #QMMM_LIBS = -L. -lmm FFLAGS = -Mr8 -tp p7 LFLAGS = -Bstatic -L. -latlas $(QMMM_LIBS) CFLAGS = CPP = /lib/cpp -P -C -traditional CPPFLAGS = -D__Linux -D__PGI -DLAPACK -DFFT_DEFAULT -DPOINTER8 -D_pgf90\ -DPARALLEL -DMP_LIBRARY=__MPI -DMYRINET NOOPT_FLAG = CC = cc FC = mpipgf90 -c -fast LD = mpipgf90 -fast AR = #---------------------------------------------------------------------------- #---------------------------------------------------------------------------- # Personal Configuration #---------------------------------------------------------------------------- SRC = /home/shyma/cpmd/P-CPMD-3.9.1/SOURCE FC = mpipgf90 -c -fast -I. -I$(SRC) ?? the problem still exists, the end part of the complied process like this "???? .o mm_mddiag.o mm_forces_diag.o mm_rho_forcedr.o vdwin.o vdw.o bs_forces_diag.o bswfo.o lsforce.o setbsstate.o wrccfl.o sysdepend.o -Bstatic -L. -latlas /usr/bin/ld: cannot find -laio make: *** [cpmd.x] Error 2 ?? so what should I do to settle the problem and complie a parallel CPMD executable. And how change the script or Makefile to suit my computer( RedHat9.0,intel4 and 2.3G, pgi5.2, lam-mpi7.0.6). I appreciate any help! Best wishes! yourse shyma -------------- next part -------------- An HTML attachment was scrubbed... URL: http://cpmd.org/pipermail/cpmd-list/attachments/20050125/e93d1ef4/attachment.html From ketajone at yahoo.com Tue Jan 25 11:18:18 2005 From: ketajone at yahoo.com (Keta Jones) Date: Tue, 25 Jan 2005 02:18:18 -0800 (PST) Subject: [CPMD-list] RMS gradient In-Reply-To: <200501061109.MAA29810@cpmd.org> Message-ID: <20050125101818.2061.qmail@web52705.mail.yahoo.com> Please help me to explain the significance of "Root mean square gradient" for energy minimization of systems.How can I check this from CPMD output file ? regards keta ===== Keta ************************************************************************************** __________________________________ Do you Yahoo!? Yahoo! Mail - You care about security. So do we. http://promotions.yahoo.com/new_mail From Ari.P.Seitsonen at iki.fi Tue Jan 25 14:20:15 2005 From: Ari.P.Seitsonen at iki.fi (Ari P Seitsonen) Date: Tue, 25 Jan 2005 14:20:15 +0100 (CET) Subject: [CPMD-list] Compiling errors, help! In-Reply-To: <41F59E19.0000BA.23879@app-01> References: <41F59E19.0000BA.23879@app-01> Message-ID: Dear Shyma, Your implementation of the MPI library has been compiled with g77 as the fortran compiler. Even though this could be made working with the pgi compiler, I'd suggest you to recompile the MPI library (mpich?) with pgi and then try again to compile CPMD. Greetings from Sunny Paris, apsi -=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=- Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/ Tel +41 1 635 44 97 / Fax +41 1 635 68 38 / GSM +41 79 719 09 35 Anschrift: Physikalisch Chemisches Institut (PCI), Universit?t Z?rich (UniZh) Indirizzo: Winterthurerstrasse 190, CH-8057 Z?rich Address: Schweiz / Svizzera / Suisse / Svizra / Switzerland On Tue, 25 Jan 2005, shyma wrote: > Dear Sir: > I'm a newer on CPMD. The version of Linux I installed was Redhat 9.0 . I have compiled a serial CPMD3.9 (the platform I chosen was PC-PGI) executable for an Intel4 processor with the pgf90, but when I tried to compiled a parallel CPMD3.9 with pgf90 (I chosen the PC-PGI-MPI platform), the error such as > " > rm -f cpmd.f > /lib/cpp -P -C -traditional -D__Linux -D__PGI -DLAPACK -DFFT_DEFAULT -DPARALLEL > -DMP_LIBRARY=__MPI -DMYRINET /home/shyma/cpmd/P-CPMD-3.9.1/SOURCE/cpmd.F ./cpmd.f > mpif77 -c -fast -I. -I/home/shyma/cpmd/P-CPMD-3.9.1/SOURCE -Mr8 -pc64 ./cpmd.f -o ./cpmd.o > g77: unrecognized option `-Mr8' > g77: unrecognized option `-pc64' > f771: unrecognized option `-fast' > mpif77: No such file or directory > make: *** [cpmd.o] Error 1 " displayed. > > TO find a solution, I change the "Makefile"(in fact i didn't understand it clearly) as fellows > "...... > # > SHELL = /bin/sh > # > #--------------- Default Configuration for pc-pgi-mpi --------------- > SRC = . > DEST = . > BIN = . > #QMMM_FLAGS = -D__QMECHCOUPL > #QMMM_LIBS = -L. -lmm > FFLAGS = -Mr8 -tp p7 > LFLAGS = -Bstatic -L. -latlas $(QMMM_LIBS) > CFLAGS = > CPP = /lib/cpp -P -C -traditional > CPPFLAGS = -D__Linux -D__PGI -DLAPACK -DFFT_DEFAULT -DPOINTER8 -D_pgf90\ > -DPARALLEL -DMP_LIBRARY=__MPI -DMYRINET > NOOPT_FLAG = > CC = cc > FC = mpipgf90 -c -fast > LD = mpipgf90 -fast > AR = > #---------------------------------------------------------------------------- > > #---------------------------------------------------------------------------- > # Personal Configuration > #---------------------------------------------------------------------------- > SRC = /home/shyma/cpmd/P-CPMD-3.9.1/SOURCE > FC = mpipgf90 -c -fast -I. -I$(SRC) > " > > the problem still exists, the end part of the complied process like this > "...... > .o mm_mddiag.o mm_forces_diag.o mm_rho_forcedr.o vdwin.o vdw.o bs_forces_diag.o > bswfo.o lsforce.o setbsstate.o wrccfl.o sysdepend.o -Bstatic -L. -latlas > /usr/bin/ld: cannot find -laio > make: *** [cpmd.x] Error 2 > " > so what should I do to settle the problem and complie a parallel CPMD executable. And how change the script or Makefile to suit my computer( RedHat9.0,intel4 and 2.3G, pgi5.2, lam-mpi7.0.6). > > I appreciate any help! > > Best wishes! > > yourse shyma From hutter at pci.unizh.ch Tue Jan 25 18:16:17 2005 From: hutter at pci.unizh.ch (Juerg Hutter) Date: Tue, 25 Jan 2005 18:16:17 +0100 (MET) Subject: [CPMD-list] RMS gradient In-Reply-To: <20050125101818.2061.qmail@web52705.mail.yahoo.com> References: <20050125101818.2061.qmail@web52705.mail.yahoo.com> Message-ID: Hi The root mean square (see http://mathworld.wolfram.com/Root-Mean-Square.html) is a measure for the size of a distribution. We use it to give a size of the electronic gradient during the minimization. The optimized energy is found for the case of zero gradient. Numerically this is monitored by the RMS value of the gradient. However, there is a second measure, the largest absolute value of an element of the gradient. In CPMD both measures are printed, side by side in the output. Convergence is checked wrt the largest gradient element. Convergence is assumed if this value falls below a threshold (default 10^-5). regards Juerg Hutter ---------------------------------------------------------- Juerg Hutter Phone : ++41 1 635 4491 Physical Chemistry Institute FAX : ++41 1 635 6838 University of Zurich E-mail: hutter at pci.unizh.ch Winterthurerstrasse 190 CH-8057 Zurich, Switzerland ---------------------------------------------------------- On Tue, 25 Jan 2005, Keta Jones wrote: > Please help me to explain the significance of "Root > mean square gradient" for energy minimization of > systems.How can I check this from CPMD output file ? > > regards > keta > > ===== > Keta > ************************************************************************************** > > > > > __________________________________ > Do you Yahoo!? > Yahoo! Mail - You care about security. So do we. > http://promotions.yahoo.com/new_mail > _______________________________________________ > CPMD-list mailing list > CPMD-list at cpmd.org > http://cpmd.org/mailman/listinfo/cpmd-list > From ketajone at yahoo.com Tue Jan 25 23:37:07 2005 From: ketajone at yahoo.com (Keta Jones) Date: Tue, 25 Jan 2005 14:37:07 -0800 (PST) Subject: [CPMD-list] RMS gradient In-Reply-To: Message-ID: <20050125223707.212.qmail@web52710.mail.yahoo.com> Dear Prof Hutter, Many many thanks for your clarification. Still I want to ask is this procedure is applicable during both the Geometry Optimization abd Molecular Dynamics scheme.If I am right, then this convergence can be checked from GNMAX/GEMAX values from output file. Suppose I want to get a optimized structure (Roughly) of pyridine dimer by Geometry Optimization(First method) and by ab Molecular Dynamics simulation(Second methos). Then can I confirm the convergence by just looking at GNMAX/GEMAX values. Which way will be better to get a optimized geometry i.e. by only OPTIMIZE GEOMETRY or MOLECULAR DYNAMIC or BOTH. Thanks again, keta --- Juerg Hutter wrote: > Hi > > The root mean square > (see > http://mathworld.wolfram.com/Root-Mean-Square.html) > is a measure for the size of a distribution. We use > it > to give a size of the electronic gradient during the > minimization. The optimized energy is found for > the case of zero gradient. Numerically this is > monitored > by the RMS value of the gradient. > However, there is a second measure, the largest > absolute > value of an element of the gradient. > In CPMD both measures are printed, side by side in > the > output. Convergence is checked wrt the largest > gradient > element. Convergence is assumed if this value > falls below a threshold (default 10^-5). > > regards > > Juerg Hutter > > ---------------------------------------------------------- > Juerg Hutter Phone : ++41 1 635 > 4491 > Physical Chemistry Institute FAX : ++41 1 635 > 6838 > University of Zurich E-mail: > hutter at pci.unizh.ch > Winterthurerstrasse 190 > CH-8057 Zurich, Switzerland > ---------------------------------------------------------- > > > On Tue, 25 Jan 2005, Keta Jones wrote: > > > Please help me to explain the significance of > "Root > > mean square gradient" for energy minimization of > > systems.How can I check this from CPMD output file > ? > > > > regards > > keta > > > > ===== > > Keta > > > ************************************************************************************** > > > > > > > > > > __________________________________ > > Do you Yahoo!? > > Yahoo! Mail - You care about security. So do we. > > http://promotions.yahoo.com/new_mail > > _______________________________________________ > > CPMD-list mailing list > > CPMD-list at cpmd.org > > http://cpmd.org/mailman/listinfo/cpmd-list > > > __________________________________ Do you Yahoo!? Yahoo! Mail - You care about security. So do we. http://promotions.yahoo.com/new_mail From hutter at pci.unizh.ch Fri Jan 28 13:14:31 2005 From: hutter at pci.unizh.ch (Juerg Hutter) Date: Fri, 28 Jan 2005 13:14:31 +0100 (MET) Subject: [CPMD-list] RMS gradient In-Reply-To: <20050125223707.212.qmail@web52710.mail.yahoo.com> References: <20050125223707.212.qmail@web52710.mail.yahoo.com> Message-ID: Hi > Dear Prof Hutter, > Many many thanks for your clarification. Still I want > to ask is this procedure is applicable during both > the Geometry Optimization abd Molecular Dynamics > scheme.If I am right, then this convergence can be > checked from GNMAX/GEMAX values from output file. > Suppose I want to get a optimized structure (Roughly) > of pyridine dimer by Geometry Optimization(First > method) and by ab Molecular Dynamics simulation(Second > methos). Then can I confirm the convergence by just > looking at GNMAX/GEMAX values. Which way will be > better to get a optimized geometry i.e. by only > OPTIMIZE GEOMETRY or MOLECULAR DYNAMIC or BOTH. If you optimize a structure with MD (using ANNEALING) you have to investigate the gradients yourself or you can add a single step of optimization and print out the gradient with the RMS and Max. abs error. regards Juerg Hutter > > Thanks again, > keta > > --- Juerg Hutter wrote: > > > Hi > > > > The root mean square > > (see > > http://mathworld.wolfram.com/Root-Mean-Square.html) > > is a measure for the size of a distribution. We use > > it > > to give a size of the electronic gradient during the > > minimization. The optimized energy is found for > > the case of zero gradient. Numerically this is > > monitored > > by the RMS value of the gradient. > > However, there is a second measure, the largest > > absolute > > value of an element of the gradient. > > In CPMD both measures are printed, side by side in > > the > > output. Convergence is checked wrt the largest > > gradient > > element. Convergence is assumed if this value > > falls below a threshold (default 10^-5). > > > > regards > > > > Juerg Hutter > > > > > ---------------------------------------------------------- > > Juerg Hutter Phone : ++41 1 635 > > 4491 > > Physical Chemistry Institute FAX : ++41 1 635 > > 6838 > > University of Zurich E-mail: > > hutter at pci.unizh.ch > > Winterthurerstrasse 190 > > CH-8057 Zurich, Switzerland > > > ---------------------------------------------------------- > > > > > > On Tue, 25 Jan 2005, Keta Jones wrote: > > > > > Please help me to explain the significance of > > "Root > > > mean square gradient" for energy minimization of > > > systems.How can I check this from CPMD output file > > ? > > > > > > regards > > > keta > > > > > > ===== > > > Keta > > > > > > ************************************************************************************** > > > > > > > > > > > > > > > __________________________________ > > > Do you Yahoo!? > > > Yahoo! Mail - You care about security. So do we. > > > http://promotions.yahoo.com/new_mail > > > _______________________________________________ > > > CPMD-list mailing list > > > CPMD-list at cpmd.org > > > http://cpmd.org/mailman/listinfo/cpmd-list > > > > > > > > > > > __________________________________ > Do you Yahoo!? > Yahoo! Mail - You care about security. So do we. > http://promotions.yahoo.com/new_mail > From msdongmo at uycdc.uninet.cm Mon Jan 31 12:13:18 2005 From: msdongmo at uycdc.uninet.cm (moise Dongmo) Date: Mon, 31 Jan 2005 10:13:18 -0100 (GMT+1) Subject: [CPMD-list] Mailing list Message-ID: <45089.195.24.196.244.1107169998.squirrel@uycdc.uninet.cm> Hi Dear sir, I want you to add me to your mailing list of CPMD package. Best regard Moise dongmo -- moise dongmo University of yaounde I Department of physics p.o.box 812 yaounde-cameroon alternate e-mail:moidong1 at yahoo.fr cellphone: 002-377-174-056 From ara_1357_2416 at yahoo.com Thu Jan 27 08:53:15 2005 From: ara_1357_2416 at yahoo.com (Younes Ansari) Date: Wed, 26 Jan 2005 23:53:15 -0800 (PST) Subject: [CPMD-list] from younes ansari Message-ID: <20050127075315.17521.qmail@web50908.mail.yahoo.com> Dear sir : I have checked out deferent methods and just got the same results. firstly I think the definitions of atoms which is in PPLIBNEW directory are deferent from other programs like Gaussian so I want to get the same result in both Gaussian and CPMD. As a result I have checked the ionization energies for Na...NA+ , K...K+ , Rb...Rb+ and figured out that the results for K and Rb are comparable with literatures but Na has a lot differences. Please tell me how I can get a Gaussian type calculation (I mean the Etot results) in the CPMD program. secondly, I have two kinds of Au,Rb,K,....in GEODECKER for example (K-q1,Kq9) which are the definitions are better to use for a simple GEOMETRY calculation. Thanks... > --------------------------------- Do you Yahoo!? Yahoo! Search presents - Jib Jab's 'Second Term' -------------- next part -------------- An HTML attachment was scrubbed... URL: http://cpmd.org/pipermail/cpmd-list/attachments/20050126/d2e80455/attachment.html