I thoroughly enjoyed the second "Promoting Female Excellence in Theoretical and Computational Chemistry" conference, both in choice of speakers and convivial atmosphere. The ratio of female/male speakers obviously favored females, but considering the nature of the conference I would not mind if it were even more skewed towards the female end of the scale, as it probably encourages audience questions from women. The format was very adequate, but ratio between established/early-career speakers might probably be adjusted slightly (for example by including a
dozen more presentations selected from poster submissions) to enable
increased "name-recognition" of younger researchers.
An
extra day or two, and some more free time for socialization would have
been very welcome: I found that the enthusiasm and conversation flow increased
substantially after the banquet talk, but by that time the meeting was
coming to an end and productive conversations had to be cut short due to
the need to catch the flights home. If the "after-banquet talk" could
be moved to the first night of the conference, the focus of
conversations during the meeting might have included more reflexion on
the sociology of our profession, the way that the subtle biases which
discourage hiring scientists with a publication-gap of a few years are
built/accepted/torn down, and so forth. That talk did serve as a wonderful
conversation starter.
I loved the presence of
children in the meeting, and think that a specific sentence in the
conference website
stating that they are welcome to the conferences would have a positive
effect in lowering barriers to attendance, and in removing the prevalent
"productivity-minded" biases which make graduate students, post-docs,
non-tenured faculty feel that embracing a scientific career must lead to
a neglect of other important parts of life. No matter how many
"empowering" talks, positive discrimination, awareness campaings, etc.,
an academic culture where powerful figures of authority (whether star
professors, PIs or funding agencies) demand or expect that researchers
put their personal life behind their scientific productivity skews the
resulting researcher pool towards the obsessively-driven,
hyper-ambitious, un-empathic tail of the population spectrum. Whether
that tail is mostly male, mostly female, or "equal-opportunity", it
favors non-collegial behavior
and chases good people away. Hyper-ambitious researchers may be very productive, but
they cannot produce much science if their behavior leads to
talented people fleeing towards other endeavors.
Congratulations to the organizing team, and a heartfelt "thank you" to all participants.
Tuesday, June 17, 2014
Thursday, April 24, 2014
Gamess (US) frequently asked questions part 6: Obtaining proper SCF convergence (Anti-)ferromagnetic coupled Fe-S clusters
Obtaining SCF convergence of FeS clusters is a very demanding task.
- obtain orbitals for bare Fe2+, Fe3+, S2-, and isolated ligands, with proper spins on the Fe atoms (5/2 for Fe3+, 2 for Fe2+)
- Manually split the "alpha/up" and "beta/down" portions of the resulting $VEC groups. For example, assuming you have a system with three Fe atoms (two Fe2+ and one Fe3+) with total spin S=5/2 and the $VEC groups for bare Fe2+ and bare Fe3+, you should cut the $VEC groups of Fe2+ and Fe3+ as:
- combine the orbitals using the small utility called combo, which you may obtain from Alex Granovsky's Firefly website.
- Manually paste the "alpha" and "beta" guesses into a single $vec group, which would be the proper guess.
- cross all your fingers and toes, and expect it to converge into the proper state. If it does not converge, change convergers (SOSCF=.T. DIIS=.F.), onset of SOSCF (SOGTOL=1e-3) , etc.
- after SCF optimization using this guess, manually scramble the ordering of Fe atoms in your input, to ascertain whether a lower energy solution can be obtained with a different spin distribution.
Good Luck!
The problem in FeS clusters is the arrangement of spins on the Fe
atoms: if you have a cluster with 4 Fe atoms, each of them with 5
up-spins, and a total spin of zero, the arrangement of spins on the
atoms could be
- Fe1 and Fe2 up-spin, Fe3 and Fe4 down-spin; or
- Fe1 and Fe4 up-spin, Fe2 and Fe3 down-spin; or
- Fe1 and Fe3 up-spin, Fe2 and Fe4 down-spin;
It goes like this:
- obtain orbitals for bare Fe2+, Fe3+, S2-, and isolated ligands, with proper spins on the Fe atoms (5/2 for Fe3+, 2 for Fe2+)
- Manually split the "alpha/up" and "beta/down" portions of the resulting $VEC groups. For example, assuming you have a system with three Fe atoms (two Fe2+ and one Fe3+) with total spin S=5/2 and the $VEC groups for bare Fe2+ and bare Fe3+, you should cut the $VEC groups of Fe2+ and Fe3+ as:
$VEC for the alpha (up) electrons of Fe2+ (let's call it "Fe2+_5_d_electrons")
$VEC for the alpha (up) electrons of Fe3+ (let's call it "Fe3+_5_d_electrons")
$VEC for the beta (down) electrons of Fe2+ (let's call it "Fe2+_1_d_electron")
$VEC for the beta (down) electrons of Fe3+ (let's call it "Fe3+_0_d_electrons")
$VEC for the alpha (up) electrons of Fe3+ (let's call it "Fe3+_5_d_electrons")
$VEC for the beta (down) electrons of Fe2+ (let's call it "Fe2+_1_d_electron")
$VEC for the beta (down) electrons of Fe3+ (let's call it "Fe3+_0_d_electrons")
The
total spin S=5/2 in this sample problem implies that both Fe2+
atoms spins should annull each other, i.e., one Fe2+ is mostly "up" and
the other is mostly "down". Building the new guess for the "up"
electrons should therefore include:
"Fe2+_5_d_electrons" for one of the Fe2+ ions,
"Fe2+_1_d_electrons" for the other Fe2+,
"Fe3+_5_d_electrons" for the Fe3+
Building the new guess for the "down" electrons should include:
"Fe2+_1_d_electrons" for the FIRST Fe2+ ions,
"Fe2+_5_d_electrons" for the other Fe2+,
"Fe3+_0_d_electrons" for the Fe3+
"Fe2+_5_d_electrons" for one of the Fe2+ ions,
"Fe2+_1_d_electrons" for the other Fe2+,
"Fe3+_5_d_electrons" for the Fe3+
Building the new guess for the "down" electrons should include:
"Fe2+_1_d_electrons" for the FIRST Fe2+ ions,
"Fe2+_5_d_electrons" for the other Fe2+,
"Fe3+_0_d_electrons" for the Fe3+
- combine the orbitals using the small utility called combo, which you may obtain from Alex Granovsky's Firefly website.
- Manually paste the "alpha" and "beta" guesses into a single $vec group, which would be the proper guess.
- cross all your fingers and toes, and expect it to converge into the proper state. If it does not converge, change convergers (SOSCF=.T. DIIS=.F.), onset of SOSCF (SOGTOL=1e-3) , etc.
- after SCF optimization using this guess, manually scramble the ordering of Fe atoms in your input, to ascertain whether a lower energy solution can be obtained with a different spin distribution.
Good Luck!
Thursday, January 16, 2014
Moving towards Open Access...
In physics and mathematics, publishing Preprints of papers in the arXiv is the most common form of distributing scientific papers. All the major journals in those areas have therefore been "forced" to accept papers previously available as preprints.
In Chemistry and Biology, however, most journals do not accept preprints and therefore authors are quite loath to make their work available as a preprint. The lack of this "free preprint" culture then enables journals to keep increasing their subscription prices way above inflation levels, which further gives publishers an extra incentive to keep rejecting sound work that might otherwise be available as costless preprints. This is a classic instance of Catch-22.
I believe that, as authors, we should do our utmost to fight this status quo. Our science should be evaluated on its merits, rather than on the accidental name of the journal where it has appeared. Therefore, I will henceforth submit all my Biochemistry work to PeerJ / PeerJPrePrints. PeerJ is an innovative and remarkably inexpensive Open Access Publisher with transparent peer-review and the option of publishing the paper's reviews alongside the manuscript. The integrity of the reviewing process is therefore above reproach, ensuring that it will be both rigorous and fair.
PeerJ does not (yet?) accept submissions outside the field of Biology. My Chemistry work must continue to be submitted elsewhere. I am thinking of given the Beilstein Journal of Organic Chemistry a shot: completely free, open access, and rigorous. It does not have a stellar IF (around 2.8, I think), but who cares? Playing the IF game is ultimately detrimental to quick publication, as several journals insist on publishing only the "extra-sexy" work to prevent their IFs from falling, and often even refuse to send manuscripts for review simply because some editor feels they are not "hot" enough (ACS, I am talking to you....)
The power to change is, after all, in our hands. It may be a very small amount of power, and the odds of effecting any change may be vanishingly small, but if we do not use it, nothing will change for sure.
In Chemistry and Biology, however, most journals do not accept preprints and therefore authors are quite loath to make their work available as a preprint. The lack of this "free preprint" culture then enables journals to keep increasing their subscription prices way above inflation levels, which further gives publishers an extra incentive to keep rejecting sound work that might otherwise be available as costless preprints. This is a classic instance of Catch-22.
I believe that, as authors, we should do our utmost to fight this status quo. Our science should be evaluated on its merits, rather than on the accidental name of the journal where it has appeared. Therefore, I will henceforth submit all my Biochemistry work to PeerJ / PeerJPrePrints. PeerJ is an innovative and remarkably inexpensive Open Access Publisher with transparent peer-review and the option of publishing the paper's reviews alongside the manuscript. The integrity of the reviewing process is therefore above reproach, ensuring that it will be both rigorous and fair.
PeerJ does not (yet?) accept submissions outside the field of Biology. My Chemistry work must continue to be submitted elsewhere. I am thinking of given the Beilstein Journal of Organic Chemistry a shot: completely free, open access, and rigorous. It does not have a stellar IF (around 2.8, I think), but who cares? Playing the IF game is ultimately detrimental to quick publication, as several journals insist on publishing only the "extra-sexy" work to prevent their IFs from falling, and often even refuse to send manuscripts for review simply because some editor feels they are not "hot" enough (ACS, I am talking to you....)
The power to change is, after all, in our hands. It may be a very small amount of power, and the odds of effecting any change may be vanishingly small, but if we do not use it, nothing will change for sure.
Wednesday, October 9, 2013
2013 Nobel Prize for Chemistry awarded to Warshel, Levitt and Karplus
This year, the Nobel Prize for Chemistry has a special flavor for computational biochemists, as the Swedish Academy has decided to recognize the creators of Quantum Mechanics/Molecular Mechanics hybrid methods. Levitt and Warshel developed the first computational simulation of a protein, back in 1975, whereas Karplus and his group developed the well-known CHARMm force-field and molecular dynamics code. I am very glad to see Warshel recognized as the towering giant he really is (in spite of his short stature). He is a very warm and likeable fellow, always eager for a good discussion on the merits (and demerits) of a scientific proposal. He does have, however, a singular lack of patience for bullshit, and a peculiar disregard for diplomatic niceties ....
Warshel's parents were Polish Jews who left Europe for Palestine before World War II to join a kibbutz. He attributes part of his peculiar temper to the circumstances of growing up in such an utopian socialist community in the 1940's, before the Israeli War of Independence. Lying and "truth embellishment" were very strongly frowned upon, and one was expected to be able to withstand withering criticism whenever warranted. His BS-detector is therefore quite over-calibrated, and he takes no prisoners when arguing against a wrong-headed argument. Warshel's abrasive style is well-known in the computational chemistry, as are is frequent controversies with other workers.
He also told me one of the best jokes I know... He presented it as a real tale of his uncle's exploits as an officer in the Polish Army. here it goes:
During on of his official leaves, his uncle went to a brothel. After the "job" had been done, he got up to get dressed and leave. The prostitute the asked: "What about the money, sir?". And he replied: "An officer never accepts money, my lady"
Warshel's parents were Polish Jews who left Europe for Palestine before World War II to join a kibbutz. He attributes part of his peculiar temper to the circumstances of growing up in such an utopian socialist community in the 1940's, before the Israeli War of Independence. Lying and "truth embellishment" were very strongly frowned upon, and one was expected to be able to withstand withering criticism whenever warranted. His BS-detector is therefore quite over-calibrated, and he takes no prisoners when arguing against a wrong-headed argument. Warshel's abrasive style is well-known in the computational chemistry, as are is frequent controversies with other workers.
He also told me one of the best jokes I know... He presented it as a real tale of his uncle's exploits as an officer in the Polish Army. here it goes:
During on of his official leaves, his uncle went to a brothel. After the "job" had been done, he got up to get dressed and leave. The prostitute the asked: "What about the money, sir?". And he replied: "An officer never accepts money, my lady"
Friday, July 5, 2013
Gamess (US) frequently asked questions Part 5: "THE VIBRATIONAL ANALYSIS IS NOT VALID"
Gamess (US) and Firefly by default assume geometric convergence has been achieved when the maximum gradient is below 1e-4 and the RMS gradient is smaller than 1/3 of the maximum gradient. This convergence criterion may be changed by the user with
$STATPT OPTTOL=<your desired convergence criterion> $END
It is well known that the vibrational analysis is strictly valid mathematically when the Hessian is computed in true stationary points (i.e when the gradient is exactly equal to zero). If the maximum gradient is sufficiently close to zero, the vibrational analysis (although not absolutely correct) is still close enough to the "true" solution for all practical purposes.
This introduction brings us to today's FAQ. A recurring question in both the Gamess-US list and the Firefly forums concerns the message often printed by the program after a vibrational analysis:
*THIS IS NOT A STATIONARY POINT ON THE MOLECULAR PES THE VIBRATIONAL ANALYSIS IS NOT VALID*
This message arises from the way gradients are analyzed by Gamess: gradients are originally computed in one set of coordinates (cartesian coordinates, I believe) , and then transformed into the coordinate system specified by the user. Optimizations stop when the "transformed gradient" lies below OPTTOL, but Gamess uses the original, non-transformed, gradient to decide whether to consider the geometry as a stationary point on the molecular PES. Therefore, if the geometry is converged, the scary message in capital letters above may be safely disregarded. When in doubt, simply decrease your OPTTOL value, continue the optimization and re-compute the hessian.
$STATPT OPTTOL=
It is well known that the vibrational analysis is strictly valid mathematically when the Hessian is computed in true stationary points (i.e when the gradient is exactly equal to zero). If the maximum gradient is sufficiently close to zero, the vibrational analysis (although not absolutely correct) is still close enough to the "true" solution for all practical purposes.
This introduction brings us to today's FAQ. A recurring question in both the Gamess-US list and the Firefly forums concerns the message often printed by the program after a vibrational analysis:
*THIS IS NOT A STATIONARY POINT ON THE MOLECULAR PES THE VIBRATIONAL ANALYSIS IS NOT VALID*
This message arises from the way gradients are analyzed by Gamess: gradients are originally computed in one set of coordinates (cartesian coordinates, I believe) , and then transformed into the coordinate system specified by the user. Optimizations stop when the "transformed gradient" lies below OPTTOL, but Gamess uses the original, non-transformed, gradient to decide whether to consider the geometry as a stationary point on the molecular PES. Therefore, if the geometry is converged, the scary message in capital letters above may be safely disregarded. When in doubt, simply decrease your OPTTOL value, continue the optimization and re-compute the hessian.
Wednesday, June 26, 2013
Gamess (US) frequently asked questions Part 4: The rungms script
Third guest post by Kirill Berezovsky (Petrozadovsk State University).
Gamess (US) is run though a script provided in the installation package. This script MUST be adapted by the user. The rungms-script below has been adapted to be used with the Gamess (US) installation described in earlier posts. Happy computing :-)
Gamess (US) is run though a script provided in the installation package. This script MUST be adapted by the user. The rungms-script below has been adapted to be used with the Gamess (US) installation described in earlier posts. Happy computing :-)
RUNGMS script
#!/bin/csh
set SCR=/scr/$USER
set USERSCR=~$USER/scr
set GMSPATH=/usr/local/gamess
set JOB=$1
set VERNO=$2
if ($VERNO == cpu) set TARGET=mpi
if ($VERNO == gpu) set TARGET=ga
if (null$VERNO == null) set VERNO=cpu
set master=`hostname`
printf "\n * Started at: `date`"
set DSK=`df -m $SCR | awk 'NR==2{print$4}'`
printf "\n * Available disk space: $DSK MB"
printf "\n * Temporary files in: $SCR and in $USERSCR\n\n"
limit stacksize 8192
if ($JOB:r.inp == $JOB) set JOB=$JOB:r
if (-e $JOB.inp) then
cp $JOB.inp
$SCR/$JOB.F05
else
echo
"Input file $JOB.inp not found"
exit 4
endif
source $GMSPATH/gms-files.csh
if (-e $HOME/.gmsrc) source $HOME/.gmsrc
set ngddi=`grep -i '^ \$GDDI' $SCR/$JOB.F05 | grep -iv
'NGROUP=0 ' | wc -l`
if ($ngddi > 0) then
set
GDDIjob=true
echo
"This is a GDDI run, keeping various output files on local disks"
set echo
setenv OUTPUT $SCR/$JOB.F06
setenv PUNCH $SCR/$JOB.F07
unset echo
else
set
GDDIjob=false
endif
if ((-e $PUNCH) || (-e $MAKEFP) || (-e $TRAJECT) ||
(-e $RESTART) ) then
echo "Please
save, rename, or erase these files from a previous run:"
echo
" $PUNCH,"
echo
" $TRAJECT,"
echo
" $RESTART, and/or"
echo
" $MAKEFP,"
echo
"and then resubmit this computation."
exit 4
endif
#----------------------------------------------------------------------
if ($TARGET == mpi) then
set NCPUS=`grep
cores /proc/cpuinfo | wc -l`
echo " * CPU
cores: $NCPUS "
echo " * GPU devices: not used"
echo " "
setenv TRAJECT $USERSCR/$JOB.trj
setenv RESTART
$USERSCR/$JOB.rst
setenv INPUT
$SCR/$JOB.F05
setenv PUNCH
$USERSCR/$JOB.dat
if ( -e
$TRAJECT ) rm $TRAJECT
if ( -e $PUNCH ) rm $PUNCH
if ( -e $RESTART ) rm $RESTART
setenv
LD_LIBRARY_PATH /opt/intel/impi/4.0.2.003/intel64/lib:$LD_LIBRARY_PATH
set path= (
/opt/intel/impi/4.0.2.003/intel64/bin $path )
mpdboot
mpiexec -n
$NCPUS $GMSPATH/gamess.$VERNO.x
mpdallexit
cp $PUNCH .
endif
#----------------------------------------------------------------------
if ($TARGET == ga) then
set PPN=1
set NCPUS=1
@ NPROCS =
$NCPUS
setenv HOSTFILE
$SCR/$JOB.nodes.mpd
if (-e
$HOSTFILE) rm $HOSTFILE
touch $HOSTFILE
echo `hostname`
>> $HOSTFILE
set NNODES=1
setenv PROCFILE
$SCR/$JOB.processes.mpd
if (-e
$PROCFILE) rm $PROCFILE
touch $PROCFILE
echo "-n
$NPROCS -host `hostname` $GMSPATH/gamess.$VERNO.x" >> $PROCFILE
set
path=(/opt/intel/impi/4.0.2.003/intel64/bin $path)
setenv
I_MPI_WAIT_MODE enable
setenv
I_MPI_PIN disable
setenv
I_MPI_DEBUG 0
setenv
I_MPI_STATS 0
setenv
I_MPI_DEVICE sock
setenv
I_MPI_NETMASK ib0
setenv
LD_LIBRARY_PATH /opt/intel/impi/4.0.2.003/intel64/lib:$LD_LIBRARY_PATH
setenv
LD_LIBRARY_PATH
/opt/intel/composerxe-2011.4.191/compiler/lib/intel64:$LD_LIBRARY_PATH
setenv
LD_LIBRARY_PATH /opt/intel/composer_xe_2013.0.079/mkl/lib/intel64:$LD_LIBRARY_PATH
setenv
MKL_SERIAL YES
setenv
MKL_NUM_THREADS 1
setenv
LD_LIBRARY_PATH /usr/local/cuda/lib64:$LD_LIBRARY_PATH
setenv
GMS_CCHEM '1'
@ NUMGPU=1
setenv CCHEM
'devices=0;memory=4g'
echo " * CPU cores: `grep cores /proc/cpuinfo | wc -l` "
echo " * GPU devices:
$NUMGPU (with settings: $CCHEM)"
echo "
"
chdir $SCR
set echo
mpdboot
--rsh=ssh -n $NNODES -f $HOSTFILE
mpiexec
-configfile $PROCFILE < /dev/null
mpdallexit
unset echo
rm -f $PROCFILE
endif
#----------------------------------------------------------------------
echo ----- accounting info -----
if ($GDDIjob == true) cp $SCR/$JOB.F07 ~/scr/$JOB.dat
echo Files used on the master node $master were:
ls -lF $SCR/$JOB.*
rm -f
$SCR/$JOB.F*
if (-e $SCR/$JOB.V84) mv $SCR/$JOB.V84 $USERSCR
if (-e $SCR/$JOB.V80) rm -f $SCR/$JOB.V*
if (-e $SCR/$JOB.TEMP02) rm -f $SCR/$JOB.TEMP*
if (-e $SCR/$JOB.orb) mv $SCR/$JOB.orb $USERSCR
if (-e $SCR/$JOB.vec) mv $SCR/$JOB.vec $USERSCR
if (-e $SCR/$JOB.mol) mv $SCR/$JOB.mol $USERSCR
if (-e $SCR/$JOB.molf) mv $SCR/$JOB.molf $USERSCR
if (-e $SCR/$JOB.mkl) mv $SCR/$JOB.mkl $USERSCR
if (-e $SCR/$JOB.xyz) mv $SCR/$JOB.xyz $USERSCR
ls $SCR/${JOB}-*.cube > $SCR/${JOB}.lis
if (! -z $SCR/${JOB}.lis) mv $SCR/${JOB}*.cube
$USERSCR
rm -f $SCR/${JOB}.lis
ls $SCR/${JOB}-*.grd > $SCR/${JOB}.lis
if (! -z $SCR/${JOB}.lis) mv $SCR/${JOB}*.grd $USERSCR
rm -f $SCR/${JOB}.lis
ls $SCR/${JOB}-*.csv > $SCR/${JOB}.lis
if (! -z $SCR/${JOB}.lis) mv $SCR/${JOB}*.csv $USERSCR
rm -f $SCR/${JOB}.lis
if ($TARGET == mpi) then
set
nnodes=`wc -l $HOSTFILE`
set
nnodes=$nnodes[1]
@ n=1
set
master=`hostname`
set
master=$master:r
while ($n
<= $nnodes)
set
host=`sed -n -e "$n p" $HOSTFILE`
set
host=$host[1]
if ($host
!= $master) then
echo
Files used on node $host were:
ssh
$host -l $USER "ls -l $SCR/$JOB.*"
ssh
$host -l $USER "rm -f $SCR/$JOB.*"
endif
@ n++
end
rm -f
$HOSTFILE
if
($?I_MPI_STATS) then
if
($I_MPI_STATS > 0) mv $SCR/stats.txt ~/$JOB.$NCPUS.stats
endif
endif
date
time
exit
Subscribe to:
Posts (Atom)