TAD questions

Dear all,

Thany you for your attention. I have some questions about the TAD when I simualted with it.

Firstly, I want to know how to estimate the correct of results obtained from MD&TAD.

Secondly, in the process of simulation, the E may become zero in sometime, I don’t whether this situation is wrong or not? but when I reset the parameters of NEB, those energy will become some positive value.

Thirdly, I want to know whether the initial configuration of system in TAD simulation is the configuration of system in the MD simulation, for example, when I introduced a vacancy into the perfect SiC, after the MD relaxed at high temperature, the configuration may become very disorder, but I am very surprised that the initial TAD configuration did not change a lot when compared with the unrelaxed MD configuration.

Here is my in.file

temperature accelerated dynamics model for vacancy clusters in sic

events occur when a neighboring atom diffuses to the vacant site

run this on multiple partitions as

mpirun -np 4 /lmp -partition 4x1 -in in.sic

change the etol from 0 to 0.0001

units metal # sets units to ‘metal’ units - Atomic units w/ ps timescale
atom_style atomic
atom_modify map array sort 0 0.0
boundary p p p
neighbor 2.0 bin # sets maximum neighbor search radius to cutoff value, using bin-sort algorithm
neigh_modify delay 0 every 10 check yes #one 10000 page 100000 # checks if neighbor list should be rebuilt after and every 5 steps

Create geometry using internal stuff

lattice diamond 4.3765
read_data data.sic
mass 1 12.0107 # type 1 = C
mass 2 28.0855 # type 2 = Si
dump 0 all custom 1 dump.init id type xu yu zu
run 0
undump 0

temperatures and velocity

variable tlo equal 2000.0
variable thi equal 2500.0
velocity all create ${thi} 52872745 mom yes rot yes dist gaussian units box

minimization parameters

variable etol equal 1.0e-5
variable ftol equal 1.0e-5
variable maxiter equal 10000
variable maxeval equal 10000
variable dmax equal 1.0e-1

Define potential

pair_style tersoff/zbl
pair_coeff * * SiC.tersoff.zbl C Si
timestep 0.0001
variable T equal elapsed*dt
#run the system dynamicaly
fix 1 all nve
fix 2 all langevin {thi} {thi} 0.0001 6994835 zero yes
thermo 100
thermo_style custom v_T elapsed dt step temp pe etotal press
dump 1 all custom 100 dump.start id type xu yu zu
run 10000
undump 1

Eliminate COM motion

velocity all zero linear
thermo_style custom v_T elapsed dt step temp pe etotal press
dump events all custom 1 dump.sic id type xu yu zu
compute event all event/displace 1.0 #one half of the first nearest neighbor
unfix 1
unfix 2
fix 1 all nvt temp {thi} {thi} 0.001

tad nsteps nevent tlo thi delta_conf tmax compute

[min etol ftol niter neval]

[neb etol_neb ftol_neb n1steps n2steps nevery]

[min_style min_style]

[neb_style min_style]

[neb_log logfile]

tad 1000000 10 {tlo} {thi} 0.05 1 event &
min {etol} {ftol} {maxiter} {maxeval} &
neb 1.0e-12 0.0 10000000 10000 20 &
min_style cg &
neb_style fire &
neb_log log.neb

Therefore, I hope you can give me some suggestions about this TAD method. Thank you very much!

I don't think your questions are LAMMPS specific, but about
TAD generally. Perhaps Aidan can give you some advice.


TAD is a complicated method that can generate correct results that are
unachievable by other means. But it is also susceptible to various kinds
of errors. I suggest reading the papers that we cite on the doc page for
more guidance on how to construct and analyze TAD simulations:

(Voter) Sørensen and Voter, J Chem Phys, 112, 9599 (2000)

(Voter2) Voter, Montalenti, Germann, Annual Review of Materials Research
32, 321 (2002).