Closed yanxon closed 4 years ago
@timfrol
Here is the MD for 300 K (D = 0.95 A^2/ps) after 1ns of equilibration:
400K (D = 4.18 A^2/ps):
Here is the new plot of Arrhenius:
@yanxon
excellent, we nailed that one. It appears that the high T point doesn't follow the perfect Arrhenius behavior. Is this a question of statistics, longer simulation time? Once the slope match the published data we can move on to GB diffusion calculations.
@timfrol
I reran the MD for 2.2ns (burn 1ns) 500K and 600K. I see this kink at around 2000 ps:
Perhaps, 500K behaves the same as 300K and 400 K, but it is taking longer to occur. I think. I'm currently running longer simulation.
Are you going to be in on Monday?
@timfrol
I ran the relaxation with this script:
clear
units metal
dimension 3
boundary p p p
atom_style atomic
read_data PdH.dump
group 1 type 1
group 2 type 2
pair_style eam/alloy
pair_coeff * * PdH_zhou.eam.alloy Pd H
compute eng all pe/atom
compute eatoms all reduce sum c_eng
reset_timestep 0
fix 1 all box/relax iso 0.0 vmax 0.001
thermo 10
thermo_style custom step pe vol lx ly lz press pxx pyy pzz c_eatoms
min_style cg
minimize 1e-25 1e-25 5000 10000
variable natoms equal "count(all)"
variable teng equal "c_eatoms"
variable length equal "lx"
variable ecoh equal "v_teng/v_natoms"
print "Total energy (eV) = ${teng};"
print "Number of atoms = ${natoms};"
print "Lattice constant (Angstoms) = ${length};"
print "Cohesive energy (eV) = ${ecoh};"
I got this plot for the E(y-axis) vs V(x-axis):
There might be phase transition!
@yanxon
Howard, can you please provide more details? Are you changing the number of H atoms? Energy is a good parameter to look at. Are you saying that there are outlires? In addition it would be good to compare the initialial number of both H and Pd with the numbers after the relaxation.
@timfrol
The number of atoms stays the same, aka 2048 Pd and 819 H. Here is the log file:
LAMMPS (10 Aug 2015)
Reading data file ...
orthogonal box = (0 0 0) to (33.3144 33.3144 33.3144)
1 by 1 by 1 MPI processor grid
reading atoms ...
2867 atoms
2048 atoms in group 1
819 atoms in group 2
WARNING: Resetting reneighboring criteria during minimization (../min.cpp:168)
Neighbor list info ...
1 neighbor list requests
update every 1 steps, delay 0 steps, check yes
master list distance cutoff = 7.35
ghost atom cutoff = 7.35
Setting up cg style minimization ...
Unit style: metal
Memory usage per processor = 3.65105 Mbytes
Step PotEng Volume Lx Ly Lz Press Pxx Pyy Pzz eatoms
0 -5342.8968 36973.849 33.314366 33.314366 33.314366 413482.7 414506.77 390085.35 435855.97 -5342.8968
10 -5605.3255 38094.193 33.64751 33.64751 33.64751 338194.9 339135.46 316548.03 358901.22 -5605.3255
20 -5823.1731 39236.944 33.980653 33.980653 33.980653 273586.35 274464.33 253496.88 292797.86 -5823.1731
.
.
.
2440 -9803.0948 34151.414 32.444137 32.444137 32.444137 0.030968224 -61.352703 -35.198474 96.644082 -9803.0948
Loop time of 27.5529 on 1 procs for 2440 steps with 2867 atoms
Minimization stats:
Stopping criterion = energy tolerance
Energy initial, next-to-last, final =
-5342.89683858 -9803.0947871 -9803.0947871
Force two-norm initial, final = 28643.4 0.00342891
Force max component initial, final = 28626.1 0.00203344
Final line search alpha, max atom move = 0.270424 0.000549891
Iterations, force evaluations = 2440 4190
Pair time (%) = 26.2664 (95.3308)
Neigh time (%) = 0.171029 (0.62073)
Comm time (%) = 0.163721 (0.594204)
Outpt time (%) = 0.020715 (0.0751826)
Other time (%) = 0.931046 (3.37912)
Nlocal: 2867 ave 2867 max 2867 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Nghost: 6769 ave 6769 max 6769 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Neighs: 189605 ave 189605 max 189605 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Total # of neighbors = 189605
Ave neighs/atom = 66.1336
Neighbor list builds = 24
Dangerous builds = 0
Total energy (eV) = -9803.09478710058;
Number of atoms = 2867;
Lattice constant (Angstoms) = 32.4441372220902;
Cohesive energy (eV) = -3.41928663658897;
So, the volume will expand during the relaxation. Then at some point, perhaps, it get to a certain volume, the atoms rearrange themselves in which lowers the formation energy.
In case you want to run it yourself. Here is the PdH.dump file and the potentials:
If you just want to run it yourself, it takes very short
@timfrol
I visualized the minimization scheme with Ovito. There is no phase transition. The minimization actually moves all the hydrogen atoms that was previously close to the Pd atoms to somewhere in between the lattice.
Do you think it's better to run the MD simulation with the minimized structure?
@yanxon
Howard, this sounds good. Perhaps you could try inserting atoms where the minimization takes them, further from the Pd atoms into interplanar spacing. In terms of the initial configuration for the MD simulation, it should not matter too much if the atoms do not fly out. This was a test for us, it maybe more important for future GB simulations.
@timfrol
I think so too. I don't think the MD result is going to change a lot. I have been wasting time on the submission on Borax these past few days. I submitted some jobs and waited for them to run. Then, it turns out there are some bugs in the script. But, I think I have fixed it. Let's see if the MD simulation with more Temperature points works.
@timfrol
I try Borax again last night. I ran the Pd bulk simulation in parts instead of fully-automation. So, I ran 300K, 325K, 350K, ... 500K for 4 hours/submission instead of 36 hours.
After minimizing the Pd structure with 40% hydrogen, I would be able to obtain closer activation value. Now, it is about 7% error. The red circles are the data before structure minimization.
I'm gonna try to run 550K and 600K to complete the data.
@yanxon Hi Howard,
I think the slope is good. Why do you think your curve is shifted upward? You are using NPT correct?
@timfrol
Both red and blue are using the same exact script (NPT). The only difference is the blue dataset ran with minimized structure.
Perhaps, the structure is not minimized, so that the hydrogen can travel more distance before a collision. Thus, this subjected to higher diffusivity.
@yanxon
There should not be any different, perhaps your equilibration stage is too short.
@yanxon
Howard, why don't we work on this together next week. I will ask you for your scripts and I will run them myself to see where the discrepancy comes from.
I will need your script that creates the structure with Pd and H, with and without minimization and the lammps script that runs the simulations and outputs the MSQD. In the meantime, you can focus on the potential development.
Hi @timfrol ,
I give you a folder in Borax. The folder contains 6 items:
When I try give frolov2 PdH_bulk/
, it gives me this error:
[yanxon1@borax2:diffusion]$ give frolov2 PdH_bulk/
Giving PdH_bulk/.
/bin/cp: omitting directory 'PdH_bulk/'
Copy of PdH_bulk/ was unsuccessful.
Removing /usr/give/frolov2/yanxon1/ from give spool.
Problem running unlink.
No files given.
@yanxon great, I will play with it next week
@timfrol
I created a tar.gz file so that I can give it to you on Borax. The filename is PdH_bulk.tar.gz.
Also, below is a more complete graph:
For higher temperature, it seems like the ln(D) agrees more with the literature. However, for the lower temperature, the ln(D) seems to have lower value.
@yanxon
Thanks Howard, I will take a look soon. Again, the biggest issue I see is that your red and blue points are different. There should not be any difference at all. There must be a bug or something. This is the first thing to check.
@timfrol
Sounds good. Thank you, Tim.
Hi @timfrol,
I hope your vacation went well.
I created this repo in an attempt to calculate the diffusivity constant.
Please check out the LAMMPS scripts.