[DFTB-Plus-User] What causes convergence errors for excited state MD?
Ben Hourahine
benjamin.hourahine at strath.ac.uk
Fri Jun 15 01:52:02 CEST 2018
Hello Sarah,
The convergence problem is actually in the ground state part of the
calculation, as this is an iterative self-consistent solution so can
become unstable. The excited state calculations, at least in this form
of linear response, is not iterative.
There are some things that might help:
* Is your initial structure geometrically relaxed?
* Why are there two types of H atom?
* Finite electronic temperature can be used in some cases (see below).
The linear response implementation can calculate excitation energies
with fractional fillings, but is not currently able to evaluate
properties of the excited state like forces or atomic charges in these
cases. If you switch off most of the options in the excited state block,
you could at least get the energy (but this cannot be used with force
evaluation, so is not possible during an MD calculation, but can be
calculated from stored structures). On the other hand, if the
temperature is meaningfully smaller than the HOMO-LUMO gap, you won't
get fractional occupations.
One minor thing to be aware of is that the excited state calculations
are at the DFTB2 level, while your ground state is a DFTB3 calculation.
My understanding is that there is usually a small shift in the
excitation energies between the DFTB2 and 3 models (but this the code
that can calculate this was not released by the group who tested this).
You can also probably make the calculation faster by using dipole and
energy window options in the excited state calculation (but not if you
need anything more than just the excitation energy). You may also be
able to use XLBOMD to make the MD faster.
Incidentally, the current parser version is 5.
Regards
Ben
On 06/06/18 23:26, Sarah Allec wrote:
> Hello,
>
> I am trying to run an MD simulation on the first excited state of
> azobenzene. My approach is to sample various initial conditions
> (coordinates and velocities) from a long (ground state) NVE simulation
> also ran with DFTB+. However, at some point, each of these trajectories
> exhibits SCC convergence errrors (or fractional fillings, which I know
> cannot be treated with linear response in DFTB+ yet). What usually
> causes such SCC errors when using the Casida block? Is there anything
> else I can do to alleviate this problem?
>
> Here is my input:
>
> Geometry = GenFormat {
> <<< "coords.gen"
> }
>
> Driver = VelocityVerlet {
> MovedAtoms = 1:-1
> Steps = 1000000
> TimeStep [fs] = 0.2
> #TimeStep = 0.005
> #KeepStationary
> Thermostat = None{}
> OutputPrefix = "w1h_md_coords"
> MDRestartFrequency = 10
> Velocities [AA/ps] = {
> 0.64481099 7.57423330 1.70718305
> -3.26628666 -2.20916913 -5.78763950
> 5.40722402 -6.75314792 5.30978241
> 5.30778618 -2.91911958 1.76799630
> -4.60343472 10.69041803 -1.35819145
> 3.21118507 -3.17116410 -6.11374944
> 5.17669335 1.08259474 2.85240779
> 0.34820665 5.39521427 6.72207259
> -0.48756735 3.81039092 -3.54931726
> -5.39068025 -1.33915960 -4.23751820
> -0.18488624 2.06218228 9.34707866
> -3.61130283 -10.55745242 -2.97583485
> -2.00052315 -0.40626965 -6.55559752
> -3.33918373 -2.81142794 -2.26809748
> 18.73727433 0.40692371 22.46695049
> 2.01021748 21.18631899 41.45998743
> -23.33885620 -12.95186387 6.03628030
> -5.75826148 6.49553187 20.40770018
> -2.89001783 -6.55138378 -24.45858968
> 25.47496930 9.98015540 1.36140634
> -7.18414811 -9.29242585 -1.58564830
> 18.33698100 0.92487779 16.46682718
> 14.19943876 -1.17927735 -18.07153124
> -1.17641620 -24.98710393 5.23523296
> }
> ConvergentForcesOnly = Yes
> KeepStationary = Yes
> #Xlbomd
> #XlbomdFast
> #Masses
> }
>
> Hamiltonian = DFTB {
>
> SCC = Yes
> SCCTolerance = 1.0e-5
> MaxSCCIterations = 5000
> SlaterKosterFiles = Type2FileNames { # File names from two atom type
> names
> Prefix =
> "/rhome/salle008/bigdata/azobenzene_last_try/trans/W1H/3ob-3-1/" # Path
> as prefix
> Separator = "-" # Dash between type names
> Suffix = ".skf" # Suffix after second type name
> }
>
> MaxAngularMomentum {
> C = "p"
> H = "s"
> N ="p"
> H1 = "s"
> }
> Filling = Fermi {
> Temperature [Kelvin] = 0
> }
> Mixer = Broyden {
> MixingParameter = 0.20000000000000001
> }
> ReadInitialCharges = No
> ForceEvaluation = "dynamics"
> DampXH = Yes
> DampXHExponent = 4.00
> ThirdOrderFull = Yes
> HubbardDerivs {
> N = -0.1535
> C = -0.1492
> H = -0.1857
> H1 = -0.1857
> }
> Dispersion = DftD3 {}
> }
> ExcitedState = Casida {
> NrOfExcitations=10
> StateOfInterest=1
> Symmetry=singlet
> WriteTransitions=yes
> WriteSPTransitions=yes
> WriteMulliken=yes
> WriteCoefficients=yes
> WriteEigenvectors=yes
> WriteTransitionDipole=yes
> }
> Options = {
> WriteAutotestTag = Yes
> WriteDetailedXML = Yes
> WriteResultsTag = Yes
> WriteDetailedOut = Yes
> RandomSeed = 0
> }
>
> ParserOptions {
> ParserVersion = 4
> }
>
> Any help is appreciated!
>
> Best,
> *Sarah Allec*
> Graduate Student
> Nanoscale & Mesoscale Energy Materials Group
> Department of Materials Science & Engineering
> University of California Riverside
> Email: sarah.allec at email.ucr.edu <mailto:sarah.allec at email.ucr.edu>
> Website: _www.bmwong-group.com <http://www.bmwong-group.com>_
>
>
> _______________________________________________
> DFTB-Plus-User mailing list
> DFTB-Plus-User at mailman.zfn.uni-bremen.de
> https://mailman.zfn.uni-bremen.de/cgi-bin/mailman/listinfo/dftb-plus-user
>
--
Dr. B. Hourahine, SUPA, Department of Physics,
University of Strathclyde, John Anderson Building,
107 Rottenrow, Glasgow G4 0NG, UK.
+44 141 548 2325, benjamin.hourahine at strath.ac.uk
2013/14 THE Awards Entrepreneurial University of the Year
2012/13 THE Awards UK University of the Year
The University of Strathclyde is a charitable body,
registered in Scotland, number SC015263
More information about the DFTB-Plus-User
mailing list