[Pw_forum] fft_scalar_MOD_cfft3d at fft_scalar.f90:1218 error in vc-relax
Dear Paolo, I will try with your settings but it will take some time for me because i have no MKL library. so after getting it, i will let you know whether it is working or not in my case. Thank you very mush for your kind help. On Mon, Apr 8, 2013 at 4:06 PM, Paolo Giannozzi wrote: > On Sun, 2013-04-07 at 19:46 +0545, Bramha Pandey wrote: > > > I am using external FFTW v.3 (-D__FFTW3 in make.sys) and FFT_LIBS > > = -lfftw3 > > I cannot reproduce it with the FFTW interface of the MKL > library. It might depend upon you fftw3 library. > > P. > > -- > Paolo Giannozzi, Dept. Chemistry, > Univ. Udine, via delle Scienze 208, 33100 Udine, Italy > Phone +39-0432-558216, fax +39-0432-558222 > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum > -- Thanks and Regards Bramha Prasad Pandey Indian School of Mines(ISM) Dhanbad, INDIA. -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20130408/a4cca28e/attachment.html
[Pw_forum] Difficulties with scf convergence
Dear Masoud, thank you very much for your reply and suggestions. I haven't tried to use nspin=2 yet, but I will see if it helps. The distance between the atoms at the interface is around 2.5 angstrom. I will let you know if your suggestions solves the problem. Thanks! BR, NV ???, 8 ?? 2013, 16:03 +02:00 ?? Masoud Nahali : >Dear ?? ?? > > >Have you tried the calculation considering spin polarization, nspin=2 and >checking the convergence ? > >you can try a smaller degauss value using marzari-vanderbilt or gaussian too. >Please check the smearing energy contribution to your total energy. > > >Are the interface atoms in close distances from each other ? I mean 1.2-2.0 >angstrom for instance. If not and you put them at far distances from each >other you may relax a van der Waals interaction in which the pure DFT is weak. > >Do you really need such high amount of cutoff for energy and density ? > >I hope it helps. > > > > > > >? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? >Best Wishes, m > >? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?? >? > > > >Masoud Nahali >SUT >masoud.nahali at gmail.com ? >alum.sharif.edu/~m_nahali ? > > > > >On Mon, Apr 8, 2013 at 12:00 PM, ?? ?? wrote: >> >>? Hi everybody, >> >>I am trying to optimize the structure of an interface between metal and >>insulator but have difficulties with scf convergence. Here are the changes in >>energies during scf cycle: >>estimated scf accuracy < 17.58555120 Ry >>estimated scf accuracy < 14.17435383 Ry >>estimated scf accuracy < 2.05765892 Ry >>estimated scf accuracy < 2.09375911 Ry >>estimated scf accuracy < 1.37675281 Ry >>estimated scf accuracy < 0.42594758 Ry >>estimated scf accuracy < 0.29104539 Ry >>estimated scf accuracy < 0.51143695 Ry >>estimated scf accuracy < 0.18765814 Ry >>estimated scf accuracy < 0.92590998 Ry >>estimated scf accuracy < 0.93126088 Ry >>estimated scf accuracy < 0.44023978 Ry >>estimated scf accuracy < 0.33756785 Ry >>estimated scf accuracy < 0.34216551 Ry >>estimated scf accuracy < 0.36573444 Ry >>estimated scf accuracy < 0.29089342 Ry >>estimated scf accuracy < 0.29287805 Ry >>estimated scf accuracy < 0.16940331 Ry >>estimated scf accuracy < 0.17503222 Ry >>estimated scf accuracy < 0.09708451 Ry >>estimated scf accuracy < 0.07751341 Ry >>estimated scf accuracy < 0.06774554 Ry >>estimated scf accuracy < 0.15834930 Ry >>estimated scf accuracy < 0.12269823 Ry >>estimated scf accuracy < 0.12679734 Ry >>estimated scf accuracy < 0.08558475 Ry >>estimated scf accuracy < 0.04411767 Ry >>estimated scf accuracy < 0.04754042 Ry >> >>I found similar posts on the forum ?but it didn't solve the problem. I >>checked my structure with xcrysen...it seems to be ok. I am using USPP. >>Any help and suggestions would be greatly appreciated. >> >>Thanks >> >>Here is my input: >> >>title = interface , >>calculation = 'relax' , >>restart_mode = 'from_scratch' , >>wf_collect = .true. , >>outdir = 'interface/' , >>pseudo_dir = './' , >>prefix = 'interface' , >>lkpoint_dir = .true. , >>disk_io = 'high' , >>verbosity = 'high' , >>nstep = 100 , >>tstress = .false. , >>tprnfor = .true. , >>/ >> >>ibrav = 8, >>celldm(1) = 9.3957, >>celldm(2) = 1.7321, >>celldm(3) = 6., >>nat = 72, >>ntyp = 3, >>ecutwfc = 50 , >>ecutrho = 500 , >>nbnd = 500, >>occupations = 'smearing' , >>degauss = 0.025 , >>smearing = 'methfessel-paxton' , >>nspin = 1 , >>/ >> >>electron_maxstep = 200, >>conv_thr = 1.0d-6 , >>mixing_mode = 'local-TF' , >>mixing_beta = 0.1 , >>mixing_ndim = 8 , >>/ >> >>ion_dynamics = 'bfgs' , >>/ >>ATOMIC_SPECIES >>? BR, >> >>N V > >___ >Pw_forum mailing list >Pw_forum at pwscf.org >http://pwscf.org/mailman/listinfo/pw_forum -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20130408/c4c217f4/attachment.html
[Pw_forum] compile as a C++ library (prototype)
Dear all, After some trial and error, following advices given here, i come up with the solution on how to make a c++ library from PW with a controllable MPI communicator. Everything compiles fine with a small example, but i did not yet implement passing input file name. In a hope that it will be useful / interesting for other users / developers, i write a quick list of how to achieve it below. Things can definitely be made more pretty, i would be grateful for any advices. Maybe a more clean version of this can make it to a release one day. I think such an interface makes it easier to combine different codes while having a full control over MPI. .check_pw.x.j runs fine so far. I did not try anything else yet. I hope the content below is appropriate for this mailing list, please, excuse me if it is not. HOWTO: 1. pwscf.f90 -- i wrapped the main function in another function. PROGRAM pwscf USE parallel_include ! INTEGER :: comm = mpi_comm_world INTEGER :: sz CHARACTER(len=256) ::name INTEGER :: do_init = 1 WRITE (*,*) ' pwscf do_init= ', do_init CALL pwscf2(comm,do_init,name,sz) END PROGRAM pwscf SUBROUTINE pwscf2 (input_comm,do_init, in_string,in_size) ? USE parallel_include ? INTEGER, INTENT (IN):: input_comm CHARACTER(len=256), INTENT (IN) :: in_string INTEGER, INTENT (IN) :: in_size INTEGER, INTENT (IN) :: do_init #ifdef __MPI ! CALL set_mp_comm(input_comm, do_init) ? 2. In parallel_include.f90 the last function in the above is defined as: ... INTEGER, PUBLIC :: qe_mp_world = mpi_comm_world LOGICAL, PUBLIC :: do_mpi_init = .TRUE. ? CONTAINS SUBROUTINE set_mp_comm(comm, do_init) IMPLICIT NONE INTEGER, INTENT(IN) :: comm INTEGER, INTENT(IN) :: do_init do_mpi_init = (do_init.EQ.1) qe_mp_world = comm RETURN END SUBROUTINE set_mp_comm 3. In mp.f90 and mp_global.f90 i replaced all the occurrence of mpi_comm_world with a new communicator qe_mp_world, which by default is equal to mpi_comm_world. In addition MPI_Init and MPI_Finalized are wrapped in IF statement, e.g. IF (do_mpi_init) CALL mpi_init(ierr). We don't want to run those if MPI was initialized already externally. 4. a c++ wrapper looks like: #include "pw.h" extern"C" { void pwscf2_(int * comm,int *do_init , char *file, int * size); } void qe_run(char *name, MPI_Comm my_comm ) { MPI_Fintfcomm; fcomm = MPI_Comm_c2f(my_comm); int size = sizeof(name); int do_init = 0; pwscf2_(, _init, name, ); } with a trivial header: #include "mpi.h" void qe_run(char *name, MPI_Comm my_comm ); The whole thing can be run as: #include "pw.h" using namespace std; int main(int argc, char* argv[]) { // Initialize the MPI library: MPI::Init(argc, argv); int me,nprocs; MPI_Comm_rank(MPI_COMM_WORLD,); MPI_Comm_size(MPI_COMM_WORLD,); char name [] = "my_input.in"; qe_run(name, MPI_COMM_WORLD); // Tell the MPI library to release all resources it is using: MPI::Finalize(); return 0; } 5. Compiling: All the c++ files are compiled in a usual way with mpic++. The wrapper-library is compiled within PW/Makefile using: MPI_LINK_FLAGS = $(shell mpic++ --showme:link) CPP_LINK_FLAGS = -L/opt/local/lib/gcc47/ -lstdc++ libpwc++.a : pw.x $(LD) -shared -fpic $(LDFLAGS) pw.o $(PWOBJS) libpw.a $(QEMODS) $(LIBOBJS) $(LIBS) $(MPI_LINK_FLAGS) $(CPP_LINK_FLAGS) -o $@ I did not yet look at how to exactly to pass the input file, but that should not be a big deal anyway. Kind regards, Denis -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20130408/62203fb5/attachment.html
[Pw_forum] compatibility between pseudopotential choice and input_dft token
Dear Cristian As usual, what is not suggested should be tested...:-) AFAIK you should do not too much harm if you use a PBE pseudopotential in a input_dft='vdwdf' calculation. If you are able to generate a rPBE pseudopotential, you may like to check differences on the same systems used in the related papers (graphite, ...). Yours G. On Monday 08 April 2013 15:55:42 Cristian Degli Esposti Boschi wrote: > >> Or a GGA (say C.pbe-mt_fhi.UPF in my case) > > > > All the vdW corrections (semiempirical DFTD2 included, london=.true., > > in ) are well tested with a given kind of xc functional. This > > is often discussed in the related papers. If you are not certain of > > > the functional recipes, look into the funct.f90 file. But this is not > > a hard limit. You may build your own functional, including vdW. I've > > tested, with PBE pseudopotentials, an hybrid+vdW functional time ago > > ('sla+pw+hse+vdw1'), with good results on molecule-surface > > interactions. > > Thanks for the answers; now my question can be more specific: > suppose I want to use a pseudopotential built constructed for PBE > (say C.pbe-mt_fhi.UPF in my case). If I set input_dft = "vdw-df", > that means "sla+pw+rpb+vdw1", > as far as I can see the gradient > correction on exchange is revPBE and not PBE (third index is 4), and > also there is no gradient correction on correlation (fourth index is 0). > Is this compatible with a PBE pseudopotential file that, if used without > vdw-df specification, would report an exchange-correlation scheme > "sla+pw+pbx+pbc" (1 4 3 4 0)? > > Or, should I set something like "sla+pw+pbx+pbc+vdw1" in order > to be consistent with the PBE pseudopotential file? > > The same question may hold for Cooper's variant (c09x instead of rpb above) > > The question is on how the code works internally; at the beginning > of the output there is a warning "Any further DFT definition will be > discarded", and I wonder what it means exactly... > > Thanks again for your time. Cristian - Article premier - Les hommes naissent et demeurent libres et ?gaux en droits. Les distinctions sociales ne peuvent ?tre fond?es que sur l'utilit? commune - Article 2 - Le but de toute association politique est la conservation des droits naturels et imprescriptibles de l'homme. Ces droits sont la libert?, la propri?t?, la s?ret? et la r?sistance ? l'oppression. Giuseppe Mattioli CNR - ISTITUTO DI STRUTTURA DELLA MATERIA v. Salaria Km 29,300 - C.P. 10 I 00015 - Monterotondo Stazione (RM) Tel + 39 06 90672836 - Fax +39 06 90672316 E-mail: ResearcherID: F-6308-2012
[Pw_forum] compatibility between pseudopotential choice and input_dft token
On Mon, 2013-04-08 at 15:55 +0200, Cristian Degli Esposti Boschi wrote: > The question is on how the code works internally; at the beginning > of the output there is a warning "Any further DFT definition will be > discarded", and I wonder what it means exactly... it means that what is specified as "input_dft" is used, what is written in the pseudopotential files is discarded. P. -- Paolo Giannozzi, Dept. Chemistry, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222
[Pw_forum] Difficulties with scf convergence
Dear *?? ?? * Have you tried the calculation considering spin polarization, nspin=2 and checking the convergence ? you can try a smaller degauss value using marzari-vanderbilt or gaussian too. Please check the smearing energy contribution to your total energy. Are the interface atoms in close distances from each other ? I mean 1.2-2.0 angstrom for instance. If not and you put them at far distances from each other you may relax a van der Waals interaction in which the pure DFT is weak. Do you really need such high amount of cutoff for energy and density ? I hope it helps. Best Wishes, m Masoud Nahali SUT masoud.nahali at gmail.com alum.sharif.edu/~m_nahali On Mon, Apr 8, 2013 at 12:00 PM, *?? ??* wrote: > > Hi everybody, > > I am trying to optimize the structure of an interface between metal and > insulator but have difficulties with scf convergence. Here are the changes > in energies during scf cycle: > estimated scf accuracy < 17.58555120 Ry > estimated scf accuracy < 14.17435383 Ry > estimated scf accuracy < 2.05765892 Ry > estimated scf accuracy < 2.09375911 Ry > estimated scf accuracy < 1.37675281 Ry > estimated scf accuracy < 0.42594758 Ry > estimated scf accuracy < 0.29104539 Ry > estimated scf accuracy < 0.51143695 Ry > estimated scf accuracy < 0.18765814 Ry > estimated scf accuracy < 0.92590998 Ry > estimated scf accuracy < 0.93126088 Ry > estimated scf accuracy < 0.44023978 Ry > estimated scf accuracy < 0.33756785 Ry > estimated scf accuracy < 0.34216551 Ry > estimated scf accuracy < 0.36573444 Ry > estimated scf accuracy < 0.29089342 Ry > estimated scf accuracy < 0.29287805 Ry > estimated scf accuracy < 0.16940331 Ry > estimated scf accuracy < 0.17503222 Ry > estimated scf accuracy < 0.09708451 Ry > estimated scf accuracy < 0.07751341 Ry > estimated scf accuracy < 0.06774554 Ry > estimated scf accuracy < 0.15834930 Ry > estimated scf accuracy < 0.12269823 Ry > estimated scf accuracy < 0.12679734 Ry > estimated scf accuracy < 0.08558475 Ry > estimated scf accuracy < 0.04411767 Ry > estimated scf accuracy < 0.04754042 Ry > > I found similar posts on the forum ?but it didn't solve the problem. I > checked my structure with xcrysen...it seems to be ok. I am using USPP. > Any help and suggestions would be greatly appreciated. > > Thanks > > Here is my input: > > title = interface , > calculation = 'relax' , > restart_mode = 'from_scratch' , > wf_collect = .true. , > outdir = 'interface/' , > pseudo_dir = './' , > prefix = 'interface' , > lkpoint_dir = .true. , > disk_io = 'high' , > verbosity = 'high' , > nstep = 100 , > tstress = .false. , > tprnfor = .true. , > / > > ibrav = 8, > celldm(1) = 9.3957, > celldm(2) = 1.7321, > celldm(3) = 6., > nat = 72, > ntyp = 3, > ecutwfc = 50 , > ecutrho = 500 , > nbnd = 500, > occupations = 'smearing' , > degauss = 0.025 , > smearing = 'methfessel-paxton' , > nspin = 1 , > / > > electron_maxstep = 200, > conv_thr = 1.0d-6 , > mixing_mode = 'local-TF' , > mixing_beta = 0.1 , > mixing_ndim = 8 , > / > > ion_dynamics = 'bfgs' , > / > ATOMIC_SPECIES > ? BR, > > N V > -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20130408/a5191cad/attachment.html
[Pw_forum] compatibility between pseudopotential choice and input_dft token
>> Or a GGA (say C.pbe-mt_fhi.UPF in my case) > All the vdW corrections (semiempirical DFTD2 included, london=.true., > in ) are well tested with a given kind of xc functional. This > is often discussed in the related papers. If you are not certain of > > the functional recipes, look into the funct.f90 file. But this is not > a hard limit. You may build your own functional, including vdW. I've > tested, with PBE pseudopotentials, an hybrid+vdW functional time ago > ('sla+pw+hse+vdw1'), with good results on molecule-surface > interactions. Thanks for the answers; now my question can be more specific: suppose I want to use a pseudopotential built constructed for PBE (say C.pbe-mt_fhi.UPF in my case). If I set input_dft = "vdw-df", that means "sla+pw+rpb+vdw1", as far as I can see the gradient correction on exchange is revPBE and not PBE (third index is 4), and also there is no gradient correction on correlation (fourth index is 0). Is this compatible with a PBE pseudopotential file that, if used without vdw-df specification, would report an exchange-correlation scheme "sla+pw+pbx+pbc" (1 4 3 4 0)? Or, should I set something like "sla+pw+pbx+pbc+vdw1" in order to be consistent with the PBE pseudopotential file? The same question may hold for Cooper's variant (c09x instead of rpb above) The question is on how the code works internally; at the beginning of the output there is a warning "Any further DFT definition will be discarded", and I wonder what it means exactly... Thanks again for your time. Cristian -- Cristian Degli Esposti Boschi CNR-IMM, Sezione di Bologna, via Gobetti, 101, 40129, Bologna, Italia tel. ++39 051 6399152, fax ++39 051 6399216 email: degliesposti -AT- bo.imm.cnr.it web: www.bo.imm.cnr.it/site/staff/personal_pages/degliesposti/
[Pw_forum] nanouse email discussion list
These discussions here are wonderful. I am so much impressed. A great place indeed. But I do not use QE, not yet, though installed it. Will start learning in about 2 months. I created an email discussion list on Yahoo Groups that is described shortly this way: Discussions on software used in nanotechnology, commercial and open source. How to use it, examples of programming. Connecting with other users and developers: Synopsys Sentaurus TCAD, Comsol Multiphysics, Nextnano, Nemo5, QuantumEspresso, Archimedes, and more... Discussing physical and engineering modeling problems, sharing code and ideas, etc. Hence, the list is not a competition at all to this one. Offers talking on somewhat broader subjects. It has publicly available archives: http://tech.groups.yahoo.com/group/nanouse/ Information about subscription is at that page. Or ask me directly. I guess some of you might be interested in. --- Zbigniew Koziol / ??? ?, PhD State University - Education-Science-Production Complex Orel, Russia http://nanolab.gu-unpk.ru/zbigniew/ Mobile: 8 910 206 4830
[Pw_forum] Difficulties with scf convergence
Hi everybody, I am trying to optimize the structure of an interface between metal and insulator but have difficulties with scf convergence. Here are the changes in energies during scf cycle: estimated scf accuracy < 17.58555120 Ry estimated scf accuracy < 14.17435383 Ry estimated scf accuracy < 2.05765892 Ry estimated scf accuracy < 2.09375911 Ry estimated scf accuracy < 1.37675281 Ry estimated scf accuracy < 0.42594758 Ry estimated scf accuracy < 0.29104539 Ry estimated scf accuracy < 0.51143695 Ry estimated scf accuracy < 0.18765814 Ry estimated scf accuracy < 0.92590998 Ry estimated scf accuracy < 0.93126088 Ry estimated scf accuracy < 0.44023978 Ry estimated scf accuracy < 0.33756785 Ry estimated scf accuracy < 0.34216551 Ry estimated scf accuracy < 0.36573444 Ry estimated scf accuracy < 0.29089342 Ry estimated scf accuracy < 0.29287805 Ry estimated scf accuracy < 0.16940331 Ry estimated scf accuracy < 0.17503222 Ry estimated scf accuracy < 0.09708451 Ry estimated scf accuracy < 0.07751341 Ry estimated scf accuracy < 0.06774554 Ry estimated scf accuracy < 0.15834930 Ry estimated scf accuracy < 0.12269823 Ry estimated scf accuracy < 0.12679734 Ry estimated scf accuracy < 0.08558475 Ry estimated scf accuracy < 0.04411767 Ry estimated scf accuracy < 0.04754042 Ry I found similar posts on the forum ?but it didn't solve the problem. I checked my structure with xcrysen...it seems to be ok. I am using USPP. Any help and suggestions would be greatly appreciated. Thanks Here is my input: title = interface , calculation = 'relax' , restart_mode = 'from_scratch' , wf_collect = .true. , outdir = 'interface/' , pseudo_dir = './' , prefix = 'interface' , lkpoint_dir = .true. , disk_io = 'high' , verbosity = 'high' , nstep = 100 , tstress = .false. , tprnfor = .true. , / ibrav = 8, celldm(1) = 9.3957, celldm(2) = 1.7321, celldm(3) = 6., nat = 72, ntyp = 3, ecutwfc = 50 , ecutrho = 500 , nbnd = 500, occupations = 'smearing' , degauss = 0.025 , smearing = 'methfessel-paxton' , nspin = 1 , / electron_maxstep = 200, conv_thr = 1.0d-6 , mixing_mode = 'local-TF' , mixing_beta = 0.1 , mixing_ndim = 8 , / ion_dynamics = 'bfgs' , / ATOMIC_SPECIES ? BR, N V -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20130408/228f4224/attachment.html
[Pw_forum] fft_scalar_MOD_cfft3d at fft_scalar.f90:1218 error in vc-relax
On Sun, 2013-04-07 at 19:46 +0545, Bramha Pandey wrote: > I am using external FFTW v.3 (-D__FFTW3 in make.sys) and FFT_LIBS > = -lfftw3 I cannot reproduce it with the FFTW interface of the MKL library. It might depend upon you fftw3 library. P. -- Paolo Giannozzi, Dept. Chemistry, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222
[Pw_forum] hexagonal structure
Dear all for cinnabar phase of HgTe how many position do we need for Te and Hg? do we need 3 position for each? -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20130408/c6321c2f/attachment.html
[Pw_forum] problem with bands.x
Dear Dr. Paolo Giannozzi Thank you for fixing this issue. After checking the make.sys file absolutely the same is found and corrected now. Now it is doing well for me. On Fri, Apr 5, 2013 at 4:23 PM, Paolo Giannozzi wrote: > On Fri, 2013-04-05 at 14:31 +0530, Krishna chaitanya wrote: > > > configure:3970: WARNING: parallel compiler mpif90 uses gfortran, but > > serial compiler ifort was detected > > configure:3972: WARNING: assuming F90=gfortran, discarding ifort > > so it's compiled with gfortran, not intel. Recently I have heard about > a potential problem with gfortran+MKL that might explain your error. > Go into the make.sys file, locate line starting with "BLAS_LIBS". > If it looks like this: > BLAS_LIBS = -lmkl_intel_lp64 -lmkl_sequential -lmkl_core > change it as follows > BLAS_LIBS = -lmkl_gf_lp64 -lmkl_sequential -lmkl_core > remove all executable (*.x) files, recompile > > P. > -- > Paolo Giannozzi, Dept. Chemistry, > Univ. Udine, via delle Scienze 208, 33100 Udine, Italy > Phone +39-0432-558216, fax +39-0432-558222 > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum > -- With Best Regards Dr. G. Krishna Chaitanya Assistant Professor School of Chemical Sciences SRTM University Nanded-431 606 India. -- next part ------ An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20130408/3c894d32/attachment.html