From baknoo at tokyo.rist.or.jp Thu Dec 12 11:54:33 2002 From: baknoo at tokyo.rist.or.jp (=?ks_c_5601-1987?B?udqz68Gk?=) Date: Thu, 12 Dec 2002 19:54:33 +0900 Subject: [Pw_forum] Fw: self-consistency criterion Message-ID: <002701c2a1cc$e295bea0$4b00a8c0@u95e9iatlagckyh> Dear users of PWSCF I am doing a structure optimization using pw.x When I restarted the calculation with "restart=.true." option, I found that the Self-Consistency Criterion "tr2" is changed to a smaller value. If I didnot make a mistake, the code want to have a stronger convergence criterion (and a more accurate force value) as the structure approaches equilibrium. However, How can I fix "tr2" value whether I restart it or not ? Could anybody please comment me about this? Thank you in advance. Noa -------------- next part -------------- An HTML attachment was scrubbed... URL: /pipermail/attachments/20021212/bdbf1b45/attachment.html From calderin at mjs4.phy.queensu.ca Thu Dec 12 20:30:41 2002 From: calderin at mjs4.phy.queensu.ca (Lazaro Calderin) Date: Thu, 12 Dec 2002 14:30:41 -0500 Subject: [Pw_forum] Unit cell and atomic position optimization Message-ID: <3DF8E3E1.4EA2CD21@mjs4.phy.queensu.ca> Hello every one! I got pwscf running on a SUN cluster a while ago and I have started doing some optimization of both unit cell parameters and atomic positions in a hexagonal system (hydroxyapatite) to compare with LCAO results. I have a problem similar to the one posted by Yanming Ma in his last message. That is: pwscf does not converge to the a given pressure. I tried increasing the mass of the unit cell to damp the oscillations between negative and positive values of the pressure but it did not work neither. I would be interested in any suggestions on how to perform relaxation of unit cell parameters and atomic positions at the same time in non-cubic systems using pwscf. I am trying to avoid the approach of fitting of total energies for different geometries to the solid equation of states. Thanks in advance, Lazaro From hwsheng at jhu.edu Fri Dec 13 06:59:12 2002 From: hwsheng at jhu.edu (HW Sheng) Date: Fri, 13 Dec 2002 00:59:12 -0500 Subject: [Pw_forum] speedup of md simulation Message-ID: <004001c2a26c$c41874e0$a567dc80@mrs.jhu.edu> Hi, Dear All. I am new to pwscf, and would greatly appreciate your help. I installed pwscf on my Linux/alpha cluster (10 dual processors), which employs LAM/MPI. I am trying to run molecular dynamics simulations as shown in example4. I found that the speedup is only a factor of 2, which is basically unacceptable. So, I want to know: (1) does this occur to you as well? (2) is this due to the fact that the code is not optimally parallelized or that the LAM/MPI is not well configured? How can I get a workaround? Thank you in advance. Howard PS: When I do k-point scf calculation, the speedup is acceptable, around a factor of 5 or so. -------------- next part -------------- An HTML attachment was scrubbed... URL: /pipermail/attachments/20021213/4d34ac4a/attachment.htm From giannozz at nest.sns.it Fri Dec 13 09:54:35 2002 From: giannozz at nest.sns.it (Paolo Giannozzi) Date: Fri, 13 Dec 2002 09:54:35 +0100 Subject: [Pw_forum] speedup of md simulation In-Reply-To: <004001c2a26c$c41874e0$a567dc80@mrs.jhu.edu> References: <004001c2a26c$c41874e0$a567dc80@mrs.jhu.edu> Message-ID: <200212130954.35637.giannozz@nest.sns.it> HI > [...] I installed pwscf on my Linux/alpha cluster (10 dual processors),which > employs LAM/MPI [...] I found that the speedup is only a factor of 2 > [...] > (2) is this due to the fact that the code is not optimally parallelized > or that the LAM/MPI is not well configured? ...or none of the above? unless your cluster has high-speed communication hardware (i.e. Myrinet, not Ethernet), you will not get a good speedup. You may even get "speedown" instead. > [...] PS: When I do k-point scf calculation, the speedup is acceptable, > around a factor of 5 or so. you mean parallelization on k points ? it requires much less communication than parallelization on plane waves. See also http://www.pwscf.org/performance.htm#parallelization Paolo -- Paolo Giannozzi e-mail: giannozz at nest.sns.it Scuola Normale Superiore Phone: +39/050509412 Piazza dei Cavalieri 7 Fax: +39/050509417, 050563513 I-56126 Pisa, Italy Office: Lab. NEST, Via della Faggiola 19 From roma at cea.fr Fri Dec 13 14:25:18 2002 From: roma at cea.fr (Guido Roma) Date: Fri, 13 Dec 2002 14:25:18 +0100 Subject: [Pw_forum] speedup of md simulation References: <004001c2a26c$c41874e0$a567dc80@mrs.jhu.edu> <200212130954.35637.giannozz@nest.sns.it> Message-ID: <3DF9DFBE.DB1C58D6@cea.fr> Paolo Giannozzi wrote: > > > you mean parallelization on k points ? it requires much less communication > than parallelization on plane waves. See also > http://www.pwscf.org/performance.htm#parallelization > Ciao Paolo, riguardo al parallelismo e di quello che avete messo in proposito sul nuovo sito pwscf: recentemente ho fatto qualche test sul cluster di alpha del CEA (dovevo presentare qualcosa ad una riunione CEA sul calcolo parallelo), ed in effetti il comportamento superscalare (ho uno speedup di ~70 con 64 processori, che peraltro non riesco MAI ad usare, per un calcolo su 72 at. di quarzo) ? spiegato bene dal fatto che sul processore alpha i Gflops diminuiscono in funzione delle onde piane (della met? passando da qualche migliaio a qualche decina di migliaia). La cosa si vede bene anche su un calcolo metallico su 16 proc in cui cambio NPOOLS da 1 a 8, prima lo spedup sale poi ridiscende (a causa della diminuzione della performance scalare). Su NEC SX6 invece, as expected, la performance cresce col numero di onde piane. Se t'interessa ho qualche grafichetto in proposito. Guido -- Guido Roma -- CEA-Saclay - DEN/DMN/SRMP Bat.520/130 Phone: [+33]-1-69085738 -- Fax: ...6867 -- Mobile: [+33]-6-20069085 From roma at cea.fr Fri Dec 13 14:32:02 2002 From: roma at cea.fr (Guido Roma) Date: Fri, 13 Dec 2002 14:32:02 +0100 Subject: [Pw_forum] speedup of md simulation References: <004001c2a26c$c41874e0$a567dc80@mrs.jhu.edu> <200212130954.35637.giannozz@nest.sns.it> <3DF9DFBE.DB1C58D6@cea.fr> Message-ID: <3DF9E152.449E8FC6@cea.fr> Sorry ... the previous mail was meant to Paolo only ... Guido -- Guido Roma -- CEA-Saclay - DEN/DMN/SRMP Bat.520/130 Phone: [+33]-1-69085738 -- Fax: ...6867 -- Mobile: [+33]-6-20069085 From hwsheng at jhu.edu Fri Dec 13 17:06:52 2002 From: hwsheng at jhu.edu (HW Sheng) Date: Fri, 13 Dec 2002 11:06:52 -0500 Subject: [Pw_forum] speedup of md simulation References: <004001c2a26c$c41874e0$a567dc80@mrs.jhu.edu> <200212130954.35637.giannozz@nest.sns.it> Message-ID: <00b201c2a2c1$a6a95be0$a567dc80@mrs.jhu.edu> Dear Paolo, Thanks for the reply. I understand that our network communication is a drag. We use 100M fast Ethernet. As to the planewave parallelization, I have tried other ab intio software package like CPMD. It seems the speedup there is much greater (around a factor of 5 on my 10 dual processor cluster). does the different performance stem from different methodologies that are utilized in those simulation packages? By the way, the new pwscf web page looks much nicer. Thanks. Howard > [...] I installed pwscf on my Linux/alpha cluster (10 dual processors),which > employs LAM/MPI [...] I found that the speedup is only a factor of 2 > [...] > (2) is this due to the fact that the code is not optimally parallelized > or that the LAM/MPI is not well configured? ...or none of the above? unless your cluster has high-speed communication hardware (i.e. Myrinet, not Ethernet), you will not get a good speedup. You may even get "speedown" instead. > [...] PS: When I do k-point scf calculation, the speedup is acceptable, > around a factor of 5 or so. you mean parallelization on k points ? it requires much less communication than parallelization on plane waves. See also http://www.pwscf.org/performance.htm#parallelization From gdking00 at hotmail.com Fri Dec 13 20:04:16 2002 From: gdking00 at hotmail.com (John Teller) Date: Fri, 13 Dec 2002 19:04:16 +0000 Subject: [Pw_forum] Unit cell and atomic position optimization Message-ID: Dear Lazaro, >I got pwscf running on a SUN cluster a while ago and I have started >doing some optimization of both unit cell parameters and atomic >positions in a hexagonal system (hydroxyapatite) to compare with LCAO >results. I have a problem similar to the one posted by Yanming Ma in >his >last message. That is: pwscf does not converge to the a given >pressure. >I tried increasing the mass of the unit cell to damp the oscillations >between negative and positive values of the pressure but it did not work >neither. I would be interested in any suggestions on how to perform >relaxation of unit cell parameters and atomic positions at the same time >in non-cubic systems using pwscf. Actually, I did not state that pwscf does not converge to a given pressure. I just mentioned that by comparing with other ab initio plane wave code (VASP or ABINIT), PWSCF may be less convient to get the EOS for non-cubic system, based on my limited knowledges on PWSCF. So I sent my questions on how to get the proper input parameters when I try to do geometry minisation. I also want to make sure what kind of MD ('cm','nm','mm')is the best to do geometry minisation. But nobody reply me except Guido From giannozz at nest.sns.it Mon Dec 16 10:33:27 2002 From: giannozz at nest.sns.it (Paolo Giannozzi) Date: Mon, 16 Dec 2002 10:33:27 +0100 Subject: [Pw_forum] speedup of md simulation In-Reply-To: <00b201c2a2c1$a6a95be0$a567dc80@mrs.jhu.edu> References: <004001c2a26c$c41874e0$a567dc80@mrs.jhu.edu> <200212130954.35637.giannozz@nest.sns.it> <00b201c2a2c1$a6a95be0$a567dc80@mrs.jhu.edu> Message-ID: <200212161033.27188.giannozz@nest.sns.it> Hi > [...] As to the planewave parallelization, I have tried other ab intio > software package like CPMD. It seems the speedup there is much greater > (around a factor of 5 on my 10 dual processor cluster). does the different > performance stem from different methodologies that are utilized in those > simulation packages? No, I don't think so. Remember that - there is an optimal number of processors for each problem - for small calculations, you may not gain anything from parallelization - if you have dual processors, you may choose whether running two processes per node with MPI parallelization, or one process per node with automatic parallelization provided by libraries. Verify that you are not doing both, or they will conflict. Paolo -- Paolo Giannozzi e-mail: giannozz at nest.sns.it Scuola Normale Superiore Phone: +39/050509412 Piazza dei Cavalieri 7 Fax: +39/050509417, 050563513 I-56126 Pisa, Italy Office: Lab. NEST, Via della Faggiola 19 From calderin at mjs5.phy.queensu.ca Sat Dec 14 22:21:31 2002 From: calderin at mjs5.phy.queensu.ca (Lazaro Calderin) Date: Sat, 14 Dec 2002 16:21:31 -0500 Subject: [Pw_forum] Unit cell and atomic position optimization Message-ID: <200212142121.QAA41848@mjs5.phy.queensu.ca> Dear Yanming, Thank you very much for your e-mail. In your previous mail to the forum you wrote: >Alternatively, I try to fix target pressure, and optimize a,b,c and atomic >positions using VCSMD. The key input parameters to realize this function for >Boron are as follows (other common parameters are not shown). > >ibrav=9, celldm(1)=8.5,celldm(2)=1.69,celldm(3)=1.00168 >nat=4,ntyp=1, >Calc='cm',iswitch=3, >amass=10.811,cmass=43.2, >nosym=.true.,order=3,press=0.0, > >But the final results are not satisfied. ... and this induced me to think that what was not satisfied was the target pressure. I am sorry I misinterpreted you. Any ways that is the kind of problem I am getting and I appreciate your answer. Now, about cm, cmass and "damping". The file INPUT_PW says the following for calc='cm': 'cm' : Parrinello-Rahman cell minimization by damped dynamics My understanding is that in Parrinello-Rahman kind of MD the "cell mass" is an inertia parameter that controls the motion of the unit cell. Therefore, one could use it to accelerated or slow down (probably damp is not the right word) the convergence of the optimization. If I remember well Andersen gives an optimum of 3/(4XPi^2)(mass of unit cell) for cmass. I tried this value for my system and larger and smaller values but I could not get pw to converge to the target pressure. I also looked for a real damping parameter but could not find it. It is likely hardwired or defined by pw (?). I will keep looking for more info. Please,if you know anything else regarding this point let me know. Thanks again and regards, Lazaro Calderin From mutombo at fzu.cz Mon Dec 16 11:07:48 2002 From: mutombo at fzu.cz (mutombo) Date: Mon, 16 Dec 2002 11:07:48 +0100 Subject: [Pw_forum] segmentation fault References: <200212142121.QAA41848@mjs5.phy.queensu.ca> Message-ID: <3DFDA5F4.CBDE7E25@fzu.cz> Dear list, I am getting this error message when I run the PWSCF code (pw1.1.2) : forrtl: severe (174): SIGSEGV, segmentation fault occurred My system is Alpha-Linux . How can I solve this? Pingo From mutombo at fzu.cz Wed Dec 18 16:40:20 2002 From: mutombo at fzu.cz (Mutombo) Date: Wed, 18 Dec 2002 16:40:20 +0100 Subject: [Pw_forum] segmentation fault Message-ID: <004001c2a6ab$cab9fc40$897ee793@fzu.cz> Dear list, I am getting this error message when I run the PWSCF code (pw1.1.2) : forrtl: severe (174): SIGSEGV, segmentation fault occurred My system is Alpha-Linux . How can I solve this? Here the input file I am using. I got it from another group using the same code. It worked for them very well. CONTO SCF : bulk electronic structure Sn/Si(111) &INPUT ibrav= 4, celldm(1)=12.5680,celldm(3)=2.5,nat=22,ntyp=3, pseudop(1)='Sn.lda', pseudop(2)='Si.vbc', pseudop(3)='H.lda', pseudo_dir = '../../../PWSCF/pseudo/', fixatom=0, isolve=1, iswitch=1, nbnd=40, degauss=0.002, ltaucry=.false., lxkcry=.true., tmp_dir='/tmp/', ecut(1) =12.0, beta(1) = 0.3, tr2 = 1.0d-16, lforce=.true., lstres=.false., output_pot='sqrtpot', / 0.000000000 -0.000000000 1.178356164 1 -0.155714495 -0.269705458 0.932688714 2 0.311428962 -0.000000000 0.932688714 2 -0.155714495 0.269705458 0.932688714 2 0.000000000 -0.000000000 0.789570385 2 -0.500000000 -0.288675135 0.859209179 2 0.500000000 0.288675135 0.859209179 2 0.000000000 -0.000000000 0.468890247 2 -0.500000000 -0.288675135 0.523201586 2 0.500000000 0.288675135 0.523201586 2 -0.338959747 -0.000000000 0.414491591 2 0.169479884 -0.293547752 0.414491591 2 0.169479884 0.293547752 0.414491591 2 -0.336726252 -0.000000000 0.085119367 2 0.168363136 -0.291613492 0.085119367 2 0.168363136 0.291613492 0.085119367 2 -0.166666667 -0.288675135 -0.006575914 2 -0.166666667 0.288675135 -0.006575914 2 0.333333333 0.000000000 -0.006575914 2 -0.166666667 -0.288675135 -0.221469463 3 -0.166666667 0.288675135 -0.221469463 3 0.333333333 0.000000000 -0.221469463 3 'Sn' 1 1 1.0 'Si' 1 2 1.0 'H' 1 3 1.0 0 3 3 1 0 0 0 Regards, Pingo --------------------------------------------------------------------------------------------------------------------------------------- Dr. Pingo Mutombo Institute of Physics Academy of Sciences of the Czech republic Cukrovarnicka 10 162 53 Praha 6 -------------- next part -------------- An HTML attachment was scrubbed... URL: /pipermail/attachments/20021218/3214396e/attachment.htm From degironc at sissa.it Wed Dec 18 18:14:58 2002 From: degironc at sissa.it (Stefano de Gironcoli) Date: Wed, 18 Dec 2002 18:14:58 +0100 Subject: [Pw_forum] segmentation fault References: <004001c2a6ab$cab9fc40$897ee793@fzu.cz> Message-ID: <3E00AD12.E9A6A4BC@sissa.it> dear Pingo, where does the program crash ? which routine ? can you trace back the problem to some extent ? is it a memory problem ? how much memory do you have on your machine ? does the problem disappear if you reduce the cutoff ? regards, Stefano de Gironcoli Mutombo wrote: > Dear list, > I am getting this error message when I run the PWSCF code (pw1.1.2) : > > forrtl: severe (174): SIGSEGV, segmentation fault occurred > > My system is Alpha-Linux . > How can I solve this? Here the input file I am using. I got it from another group > using the same code. It worked for them very well. CONTO SCF : bulk electronic structure > > Sn/Si(111) > &INPUT ibrav= 4, celldm(1)=12.5680,celldm(3)=2.5,nat=22,ntyp=3, > pseudop(1)='Sn.lda', > pseudop(2)='Si.vbc', > pseudop(3)='H.lda', > pseudo_dir = '../../../PWSCF/pseudo/', > fixatom=0, > isolve=1, > iswitch=1, > nbnd=40, > degauss=0.002, > ltaucry=.false., > lxkcry=.true., > tmp_dir='/tmp/', > ecut(1) =12.0, > beta(1) = 0.3, > tr2 = 1.0d-16, > lforce=.true., lstres=.false., > output_pot='sqrtpot', > / > 0.000000000 -0.000000000 1.178356164 1 > -0.155714495 -0.269705458 0.932688714 2 > 0.311428962 -0.000000000 0.932688714 2 > -0.155714495 0.269705458 0.932688714 2 > 0.000000000 -0.000000000 0.789570385 2 > -0.500000000 -0.288675135 0.859209179 2 > 0.500000000 0.288675135 0.859209179 2 > 0.000000000 -0.000000000 0.468890247 2 > -0.500000000 -0.288675135 0.523201586 2 > 0.500000000 0.288675135 0.523201586 2 > -0.338959747 -0.000000000 0.414491591 2 > 0.169479884 -0.293547752 0.414491591 2 > 0.169479884 0.293547752 0.414491591 2 > -0.336726252 -0.000000000 0.085119367 2 > 0.168363136 -0.291613492 0.085119367 2 > 0.168363136 0.291613492 0.085119367 2 > -0.166666667 -0.288675135 -0.006575914 2 > -0.166666667 0.288675135 -0.006575914 2 > 0.333333333 0.000000000 -0.006575914 2 > -0.166666667 -0.288675135 -0.221469463 3 > -0.166666667 0.288675135 -0.221469463 3 > 0.333333333 0.000000000 -0.221469463 3 > 'Sn' 1 1 1.0 > 'Si' 1 2 1.0 > 'H' 1 3 1.0 > 0 > 3 3 1 0 0 0Regards, > Pingo--------------------------------------------------------------------------------------------------------------------------------------- > > Dr. Pingo Mutombo > Institute of Physics > Academy of Sciences of the Czech republic > Cukrovarnicka 10 > 162 53 Praha 6 -------------- next part -------------- An HTML attachment was scrubbed... URL: /pipermail/attachments/20021218/014b0067/attachment.htm From gianni.profeta at aquila.infn.it Wed Dec 18 17:28:31 2002 From: gianni.profeta at aquila.infn.it (Gianni Profeta) Date: Wed, 18 Dec 2002 17:28:31 +0100 Subject: [Pw_forum] segmentation fault References: <004001c2a6ab$cab9fc40$897ee793@fzu.cz> <3E00AD12.E9A6A4BC@sissa.it> Message-ID: <3E00A22F.2604D8ED@aquila.infn.it> Dear Mutombo, the problem is related to the dxml or cxml (on Alpha). Because the problem is located in the lapack routines, try to link with -lblas -llapack before the -lcxml (or -ldxml). Gianni Profeta Stefano de Gironcoli wrote: > > dear Pingo, > > where does the program crash ? > which routine ? can you trace back the problem to some extent > ? > is it a memory problem ? how much memory do you have on your > machine ? > does the problem disappear if you reduce the cutoff ? > > regards, > Stefano de Gironcoli > > Mutombo wrote: > > > Dear list, > > I am getting this error message when I run the PWSCF code (pw1.1.2) > > : > > > > forrtl: severe (174): SIGSEGV, segmentation fault occurred > > > > My system is Alpha-Linux . > > How can I solve this? Here the input file I am using. I got it > > from another group using the same code. It worked for them very > > well. CONTO SCF : bulk electronic structure > > Sn/Si(111) > > &INPUT ibrav= 4, celldm(1)=12.5680,celldm(3)=2.5,nat=22,ntyp=3, > > > > pseudop(1)='Sn.lda', > > pseudop(2)='Si.vbc', > > pseudop(3)='H.lda', > > pseudo_dir = '../../../PWSCF/pseudo/', > > fixatom=0, > > isolve=1, > > iswitch=1, > > nbnd=40, > > degauss=0.002, > > ltaucry=.false., > > lxkcry=.true., > > tmp_dir='/tmp/', > > ecut(1) =12.0, > > beta(1) = 0.3, > > tr2 = 1.0d-16, > > lforce=.true., lstres=.false., > > output_pot='sqrtpot', > > / > > 0.000000000 -0.000000000 1.178356164 1 > > -0.155714495 -0.269705458 0.932688714 2 > > 0.311428962 -0.000000000 0.932688714 2 > > -0.155714495 0.269705458 0.932688714 2 > > 0.000000000 -0.000000000 0.789570385 2 > > -0.500000000 -0.288675135 0.859209179 2 > > 0.500000000 0.288675135 0.859209179 2 > > 0.000000000 -0.000000000 0.468890247 2 > > -0.500000000 -0.288675135 0.523201586 2 > > 0.500000000 0.288675135 0.523201586 2 > > -0.338959747 -0.000000000 0.414491591 2 > > 0.169479884 -0.293547752 0.414491591 2 > > 0.169479884 0.293547752 0.414491591 2 > > -0.336726252 -0.000000000 0.085119367 2 > > 0.168363136 -0.291613492 0.085119367 2 > > 0.168363136 0.291613492 0.085119367 2 > > -0.166666667 -0.288675135 -0.006575914 2 > > -0.166666667 0.288675135 -0.006575914 2 > > 0.333333333 0.000000000 -0.006575914 2 > > -0.166666667 -0.288675135 -0.221469463 3 > > -0.166666667 0.288675135 -0.221469463 3 > > 0.333333333 0.000000000 -0.221469463 3 > > 'Sn' 1 1 1.0 > > 'Si' 1 2 1.0 > > 'H' 1 3 1.0 > > 0 > > 3 3 1 0 0 0Regards, > > Pingo--------------------------------------------------------------------------------------------------------------------------------------- > > > > Dr. Pingo Mutombo > > Institute of Physics > > Academy of Sciences of the Czech republic > > Cukrovarnicka 10 > > 162 53 Praha 6 -------------- next part -------------- A non-text attachment was scrubbed... Name: gianni.profeta.vcf Type: text/x-vcard Size: 342 bytes Desc: Card for Gianni Profeta Url : /pipermail/attachments/20021218/715edb45/attachment.vcf From mutombo at fzu.cz Wed Dec 18 20:32:12 2002 From: mutombo at fzu.cz (Pingo Mutombo) Date: Wed, 18 Dec 2002 20:32:12 +0100 (MET) Subject: [Pw_forum] segmentation fault In-Reply-To: <3E00AD12.E9A6A4BC@sissa.it> Message-ID: Dear Stefano and Gianni, I am actually outside my institute and I cannot login to my machine to make the change you suggested.I will do it once I returned back from my vacation and I will let you know about any eventual success. Many thanks for your assistance. Regards, Pingo On Wed, 18 Dec 2002, Stefano de Gironcoli wrote: > dear Pingo, > > where does the program crash ? > which routine ? can you trace back the problem to some extent ? > is it a memory problem ? how much memory do you have on your machine ? > does the problem disappear if you reduce the cutoff ? > > regards, > Stefano de Gironcoli > > Mutombo wrote: > > > Dear list, > > I am getting this error message when I run the PWSCF code (pw1.1.2) : > > > > forrtl: severe (174): SIGSEGV, segmentation fault occurred > > > > My system is Alpha-Linux . > > How can I solve this? Here the input file I am using. I got it from another group > > using the same code. It worked for them very well. CONTO SCF : bulk electronic structure > > > > Sn/Si(111) > > &INPUT ibrav= 4, celldm(1)=12.5680,celldm(3)=2.5,nat=22,ntyp=3, > > pseudop(1)='Sn.lda', > > pseudop(2)='Si.vbc', > > pseudop(3)='H.lda', > > pseudo_dir = '../../../PWSCF/pseudo/', > > fixatom=0, > > isolve=1, > > iswitch=1, > > nbnd=40, > > degauss=0.002, > > ltaucry=.false., > > lxkcry=.true., > > tmp_dir='/tmp/', > > ecut(1) =12.0, > > beta(1) = 0.3, > > tr2 = 1.0d-16, > > lforce=.true., lstres=.false., > > output_pot='sqrtpot', > > / > > 0.000000000 -0.000000000 1.178356164 1 > > -0.155714495 -0.269705458 0.932688714 2 > > 0.311428962 -0.000000000 0.932688714 2 > > -0.155714495 0.269705458 0.932688714 2 > > 0.000000000 -0.000000000 0.789570385 2 > > -0.500000000 -0.288675135 0.859209179 2 > > 0.500000000 0.288675135 0.859209179 2 > > 0.000000000 -0.000000000 0.468890247 2 > > -0.500000000 -0.288675135 0.523201586 2 > > 0.500000000 0.288675135 0.523201586 2 > > -0.338959747 -0.000000000 0.414491591 2 > > 0.169479884 -0.293547752 0.414491591 2 > > 0.169479884 0.293547752 0.414491591 2 > > -0.336726252 -0.000000000 0.085119367 2 > > 0.168363136 -0.291613492 0.085119367 2 > > 0.168363136 0.291613492 0.085119367 2 > > -0.166666667 -0.288675135 -0.006575914 2 > > -0.166666667 0.288675135 -0.006575914 2 > > 0.333333333 0.000000000 -0.006575914 2 > > -0.166666667 -0.288675135 -0.221469463 3 > > -0.166666667 0.288675135 -0.221469463 3 > > 0.333333333 0.000000000 -0.221469463 3 > > 'Sn' 1 1 1.0 > > 'Si' 1 2 1.0 > > 'H' 1 3 1.0 > > 0 > > 3 3 1 0 0 0Regards, > > Pingo--------------------------------------------------------------------------------------------------------------------------------------- > > > > Dr. Pingo Mutombo > > Institute of Physics > > Academy of Sciences of the Czech republic > > Cukrovarnicka 10 > > 162 53 Praha 6 > From calderin at mjs4.phy.queensu.ca Wed Dec 18 23:21:00 2002 From: calderin at mjs4.phy.queensu.ca (Lazaro Calderin) Date: Wed, 18 Dec 2002 17:21:00 -0500 Subject: [Pw_forum] nmix Message-ID: <3E00F4CC.2196C3D4@mjs4.phy.queensu.ca> Dear all, I think there is a problem with pw and the nmix parameter. The troubleshooting section of the manual says the following about problems of convergence of the scf cycle: Self-consistency is slow or does not converge. Reduce the beta parameter from the default value (0.7) to 0.3 - 0.1 ... You may also try to increase nmix to more than 4 (default value)... but when you try to increase nmix pw stops with the error "Wrong nmix" because nmix is hardwired to be not greater than 4 in the program. Specifically in the file readin.F90 there is the following line: if (nmix.le.0 .or. nmix.gt.4)& & call error('readin','Wrong nmix',1) This seems to be a bug (?). Any comment would be appreciated. Thanks, Lazaro From giannozz at nest.sns.it Thu Dec 19 09:21:26 2002 From: giannozz at nest.sns.it (Paolo Giannozzi) Date: Thu, 19 Dec 2002 09:21:26 +0100 Subject: [Pw_forum] nmix In-Reply-To: <3E00F4CC.2196C3D4@mjs4.phy.queensu.ca> References: <3E00F4CC.2196C3D4@mjs4.phy.queensu.ca> Message-ID: <200212190921.26303.giannozz@nest.sns.it> Hi > [...] when you try to increase nmix pw stops with the error "Wrong nmix" > because nmix is hardwired to be not greater than 4 in the program. > Specifically in the file readin.F90 there is the following line: > > if (nmix.le.0 .or. nmix.gt.4)& > & call error('readin','Wrong nmix',1) > > This seems to be a bug (?) you can remove this test. Then you have to increase the value of parameter "maxter" in mix_pot.F90, otherwise the program will stop there complaining that "n_iter is too big"!!! In the forthcoming new release of PWscf, there is a different mixing mechanism that is much better than the previous one, at least in the cases I tried. Paolo -- Paolo Giannozzi e-mail: giannozz at nest.sns.it Scuola Normale Superiore Phone: +39/050509412 Piazza dei Cavalieri 7 Fax: +39/050509417, 050563513 I-56126 Pisa, Italy Office: Lab. NEST, Via della Faggiola 19 From yelinhui1 at yahoo.com Thu Dec 19 09:58:42 2002 From: yelinhui1 at yahoo.com (fdsa fda) Date: Thu, 19 Dec 2002 00:58:42 -0800 (PST) Subject: [Pw_forum] Phonon calculation, but inexact atomic coordinates Message-ID: <20021219085842.11858.qmail@web20703.mail.yahoo.com> Dear all, I'm trying to apply phonon calculation for a difficult case, which are carbon nanotubes. Before I begin, there is a question on the accuracy of the expected result: To my understanding, when applying the linear response method to calculate purtabations like phonon, it is crucial that the EXACT atomic coordinates on unperturbed geometries are determined first by structural relaxation run. Am I right? But to my experiences, for systems like carbon nanotubes(several tens of atoms in unit cell), the accuracy of atomic coordinates can be determined no better than about 0.01 Angstrom by structural relaxation run. Is it justified to make such an estimation? So, suppose we already have an error bar of 0.01 Angstrom on atomic positions in UNPERTURBED geometries of carbon nanotubes, then how large will the error be on the final phonon frequencies by PWSCF, or any other DFPT calculations, caused by such inaccuracy on atomic coordinates? P.S. for high symmetry crystals such errors do not happend, because atoms there all have fractional coordinates. Any reply is appreciated. Linhui __________________________________________________ Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now. http://mailplus.yahoo.com From degironc at sissa.it Thu Dec 19 09:56:33 2002 From: degironc at sissa.it (Stefano de Gironcoli) Date: Thu, 19 Dec 2002 09:56:33 +0100 Subject: [Pw_forum] Phonon calculation, but inexact atomic coordinates References: <20021219085842.11858.qmail@web20703.mail.yahoo.com> Message-ID: <3E0189C1.6E0A84FE@sissa.it> DFPT compute the phonon frequency from the curfature of the total energy with respect to atomic positions. The fact that the atomic configuration is not exactly the equilibrium one will modify the curvature but should not induce additional errors. How much the frequency are modified ? I think that one can estimate the effect from gruneisen parameters which are tipically of order unit... that is a change (or error) in volume of 1% will induce a change in frequency of 1% ... or converting this number in terms of bond-lengths I would say that a change in bonding distance by 1% would give a produce a change in the frequency by 3%. Therefore, provided the relaxation run determine the equilibrium structural properties within a fraction of % I would be rather confident on the resulting phonon frequencies. regards Stefano de Gironcoli dsa fda wrote: > Dear all, > > I'm trying to apply phonon calculation for a difficult > case, which are carbon nanotubes. Before I begin, > there is a question on the accuracy of the expected > result: > > To my understanding, when applying the linear response > method to calculate purtabations like phonon, it is > crucial that the EXACT atomic coordinates on > unperturbed geometries are determined first by > structural relaxation run. Am I right? > > But to my experiences, for systems like carbon > nanotubes(several tens of atoms in unit cell), the > accuracy of atomic coordinates can be determined no > better than about 0.01 Angstrom by structural > relaxation run. Is it justified to make such an > estimation? > > So, suppose we already have an error bar of 0.01 > Angstrom on atomic positions in UNPERTURBED geometries > of carbon nanotubes, then how large will the error be > on the final phonon frequencies by PWSCF, or any > other DFPT calculations, caused by such inaccuracy on > atomic coordinates? > > P.S. for high symmetry crystals such errors do not > happend, because atoms there all have fractional > coordinates. > > Any reply is appreciated. > > Linhui > > __________________________________________________ > Do you Yahoo!? > Yahoo! Mail Plus - Powerful. Affordable. Sign up now. > http://mailplus.yahoo.com > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum From yelinhui1 at yahoo.com Thu Dec 19 12:14:27 2002 From: yelinhui1 at yahoo.com (fdsa fda) Date: Thu, 19 Dec 2002 03:14:27 -0800 (PST) Subject: [Pw_forum] Phonon calculation, but inexact atomic coordinates In-Reply-To: <3E0189C1.6E0A84FE@sissa.it> Message-ID: <20021219111427.10009.qmail@web20708.mail.yahoo.com> Dear Stefano, Thank you very much for your kind reply. But, as you said, > DFPT compute the phonon frequency from the curfature > of the total energy with respect to atomic > positions. The fact that the atomic configuration is > not exactly the equilibrium one will modify the > curvature but should not induce additional errors. > I'm sorry to find the above words hard to understand. Did you mean, even if the input atomic coordinates are not exactly the "real" equilibrium values, errors in atomic positions will not cause any extra frequencies error, besides the "standard" DFT error(e.g., k sampling, finite cutoff energy, single particle scheme...)? Or inexact atomic positions in UNPERTURBed geometry DO cause errors in calculated frequencies? Let me raise up the question in another way. As I know, errors in lattice constants will affect the accuracy of freqencies, because modification on lattice constants means applying external pressure. I think errors caused by inexact lattice constants are easy to understand. But, if the lattice constants are correct, but atomic positions in the unit cell has some errors(just like the case in carbon nanotubes), do this kind of errors cause inaccuracy in frequencies, and, if yes(which I believe), how serious will the error be? Best wishes, Linhui __________________________________________________ Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now. http://mailplus.yahoo.com From degironc at sissa.it Thu Dec 19 12:25:15 2002 From: degironc at sissa.it (Stefano de Gironcoli) Date: Thu, 19 Dec 2002 12:25:15 +0100 Subject: [Pw_forum] Phonon calculation, but inexact atomic coordinates References: <20021219111427.10009.qmail@web20708.mail.yahoo.com> Message-ID: <3E01AC9B.DA813DC0@sissa.it> fdsa fda wrote: > > DFPT compute the phonon frequency from the curfature > > of the total energy with respect to atomic > > positions. The fact that the atomic configuration is > > not exactly the equilibrium one will modify the > > curvature but should not induce additional errors. > > > > I'm sorry to find the above words hard to understand. I wanted to say that the presence of non zero forces on the atoms in a non equilibrium structure do not create a problem in the calculation of the frequencies. Obviously if the the position are slightly different from the equilibrium ones there will be an effect on the frequencies. I would expect that the change in the frequencies due to "random" 1 % stretching (or compression) of the bonds will be of the order of what one obtains by stretching coherently all the bonds when expanding the cell. Therefore of the order of a few %. There might be some frequencies that are determined by some local environment which will be more sensittive to somespecific atomic positions but as a rule of tumb I would expect that an estimate based on the gruneisen parameters should be qualitatively correct. regards stefano From yelinhui1 at yahoo.com Thu Dec 19 13:55:06 2002 From: yelinhui1 at yahoo.com (fdsa fda) Date: Thu, 19 Dec 2002 04:55:06 -0800 (PST) Subject: [Pw_forum] Phonon calculation, but inexact atomic coordinates In-Reply-To: <3E01AC9B.DA813DC0@sissa.it> Message-ID: <20021219125506.3552.qmail@web20702.mail.yahoo.com> Dear Stefano, This time I've got it. Thank you very much! Best wishes, Linhui > I wanted to say that the presence of non zero forces > on the atoms > in a non equilibrium structure do not create a > problem in the > calculation of the frequencies. > > Obviously if the the position are slightly different > from > the equilibrium ones there will be an effect on the > frequencies. > > I would expect that the change in the frequencies > due to "random" > 1 % stretching (or compression) of the bonds will > be of the order of > what one obtains by stretching coherently all the > bonds when > expanding the cell. Therefore of the order of a few > %. > There might be some frequencies that are determined > by some local > environment which will be more sensittive to > somespecific atomic > positions but as a rule of tumb I would expect that > an estimate based > on the gruneisen parameters should be qualitatively > correct. > > regards > > stefano > > > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum __________________________________________________ Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now. http://mailplus.yahoo.com From calderin at mjs4.phy.queensu.ca Thu Dec 19 18:37:15 2002 From: calderin at mjs4.phy.queensu.ca (Lazaro Calderin) Date: Thu, 19 Dec 2002 12:37:15 -0500 Subject: [Pw_forum] nmix problem Message-ID: <3E0203CB.2CFB2046@mjs4.phy.queensu.ca> Dear all, I think there is a problem with pw and the nmix parameter. The troubleshooting section of the manual says the following about problems of convergence of the scf cycle: Self-consistency is slow or does not converge. Reduce the beta parameter from the default value (0.7) to 0.3 - 0.1 ... You may also try to increase nmix to more than 4 (default value)... but when you try to increase nmix pw stops with the error "Wrong nmix" because nmix is hardwired to be not greater than 4 in the program. Specifically in the file readin.F90 there is the following line: if (nmix.le.0 .or. nmix.gt.4)& & call error('readin','Wrong nmix',1) This seems to be an small bug. Any comment would be appreciated. Thanks, Lazaro From d_bahmed at hotmail.com Mon Dec 30 12:58:40 2002 From: d_bahmed at hotmail.com (bahmed daoudi) Date: Mon, 30 Dec 2002 12:58:40 +0100 Subject: [Pw_forum] pwscf use for allows Message-ID: An HTML attachment was scrubbed... URL: /pipermail/attachments/20021230/995b191a/attachment.htm From bth20 at cam.ac.uk Mon Dec 23 12:09:40 2002 From: bth20 at cam.ac.uk (Ben Hope) Date: Mon, 23 Dec 2002 11:09:40 -0000 Subject: [Pw_forum] Fermi energy In-Reply-To: <200412201634.40244.giannozz@nest.sns.it> Message-ID: Simple one this: I have run both scf and nscf calculations (in order to obtain a band structure). The Fermi energy does not seem to be included in the output (unless I am being very blind again). How does one obtain this figure? Regards Ben