[DFTB-Plus-User] Lattice Optimization in DFTb+
Hargreaves, Laura
laura.hargreaves.14 at ucl.ac.uk
Mon Jan 21 19:13:12 CET 2019
Apologies for the formatting, I have attached a PDF of my results.
On 21 Jan 2019, at 18:08, Hargreaves, Laura <laura.hargreaves.14 at ucl.ac.uk<mailto:laura.hargreaves.14 at ucl.ac.uk>> wrote:
Hello,
I am trying to run a lattice optimisation with bulk Titania using the tiorg-0-1 set. I have found the results and minima found depend on which optimiser I use. The following table shows my results using different constraints: free to relax, isotropic and fix angles.
The final result in all cases where the system was allowed to fully relax with no constraints led to a system which was not exactly Tetragonal, which I have found with the settings that are reported in the following paper: (1) Dolgonos, G.; Aradi, B.; Moreira, N. H.; Frauenheim, T. An Improved Self-Consistent-Charge Density-Functional Tight-Binding (SCC-DFTB) Set of Parameters for Simulation of Bulk and Molecular Systems Involving Titanium. J. Chem. Theory Comput. 2010, 6 (1), 266–278.
In each case I used the same starting geometry, (which had previously been generated using DFT therefore I don’t believe it is a poor starting geometry), a k point grid 4 X 4 X 8, SCC tolerance 1E-7, and force tolerance of 1E-4 (unless otherwise stated).
K Point size and optimizer
a /Å
b/Å
c/Å
c/a
Equatorial / Å
Axial /Å
Total Energy /eV
Band Gap /eV
4 4 8 LBFGS Free
4.633
4.634
2.997
2.997/4.633 = 0.6469
2.00
2.00
1.94
1.94
1.94
-425.5990
3.12
4 4 8 LBFGS
Angles fixed to 90 degrees
4.607
4.607
2.998
2.998/4.633 = 0.6471
2.00
2.00
1.94
1.94
-425.5713
3.11
4 4 8 Fix angles LBFGS
Angles fixed to 90 degrees Force tolerance 1E-6
4.632
4.635
2.997
0.6466
2.00
1.99
2.00
1.92
2.00
-425.5990
3.12
4 4 8 Conjugate Gradient
4.672
4.672
2.997
2.997/4.672 = 0.6415
1.99
1.99
1.99
1.99
1.99
-425.6037
3.13
4 4 8 Conjugate Gradient
Constraint used: Isotropic
4.643
4.643
2.995
2.995/4.643 = 0.6451
1.99
2.00
2.00
1.93
1.93
-425.5972
3.11
4 4 8 CG
Angles fixed to 90 degrees
4.180
4.180
4.176
0.999
3.46
3.46
1.81
1.81
1.81
-425.4453
3.928
4 4 8 gDIIS
4.672
4.672
2.997
2.997/4.672 = 0.6415
1.99
2.00
1.93
2.00
1.92
-425.6037
3.13
Although the differences are small, to build slabs requires having the best lattice parameters that I can generate, therefore, I do not want to double any error that might occur depending on which optimiser I use.
The following cases could not converge or showed an error:
LBFGS Isotropic - could not optimise by scaling the relative lattice vector
gDIIS: could not converge either by fixing angles or using the isotropic setting
My main concerns are:
* Optimisers find different minimas - could there be a bug in the optimizers?
* The axial and equatorial bond lengths can vary upto 0.08 /Å, this can lead to problems when building slabs. I have further found that when building 110 slabs, then optimising the geometry, there is a bond length difference is still dependent on which optimiser I use
* The lowest energy system (Conjugate gradient free to relax) has lattice vectors which are longer than values in comparison, to other results generated and are in worse agreement with experiment and DFT results . I know the differences between all the systems are very small, however, these results slightly worry me to what optimiser I should be using when studying further properties.
I hope this makes sense.
Many thanks,
Laura
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20190121/2d8687be/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: table.pdf
Type: application/pdf
Size: 39510 bytes
Desc: table.pdf
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20190121/2d8687be/attachment-0001.pdf>
More information about the DFTB-Plus-User
mailing list