Hi all, I have encountered a puzzling problem when running siesta calculations for a graphene arm-chair nano-ribbon, with width of 5 C atoms (basically 2 benzene rings), terminated by H atoms on the sides. The calculation is thus supposed to be 1D, periodic in Z-direction. The system has 14 atoms in rectangular (cartesian) lattice cell.
To make sure the cell size is big enough in x and y dimension for the 1D calculation, I calculated total energies of an unrelaxed system (with fixed coordinates) with different cell dimensions, 41 kpoints in z directions, and mesh cut off of 900 Ry (these are converged values, more on the mesh below). The basis I am using is SZ. I need to use the minimum basis to get result that could be used to compare with results from another tight-binding code. When scanning the output files, I did not notice any warnings or errors in the basis set, and maximum cut-off for automatically generated H and C orbitals are both around 5.1 Bohr. The basis settings are given by: PAO.BasisSize SZ PAO.EnergyShift 200 meV Here are the results: For x cell dimension, with y dimension fixed at 10 Ang. (the nano-ribbon is flat in y-plane, so I assumed 10 Ang would be enough as an initial guess) # Cell x dim (Ang) | Total Energy (eV) 10.000000 -1594.828815 15.000000 -1594.855952 20.000000 -1594.856570 25.000000 -1594.856598 30.000000 -1594.856603 35.000000 -1594.856581 40.000000 -1594.856600 45.000000 -1594.856593 50.000000 -1594.856601 60.000000 -1594.856596 70.000000 -1594.856608 And for y cell dimension, with x dimension fixed to be 30.0 Ang: # Cell y dim (Ang) | Total Energy (eV) 5.000000 -1594.558893 10.000000 -1594.856603 15.000000 -1594.860901 20.000000 -1594.861861 25.000000 -1594.862136 30.000000 -1594.862229 40.000000 -1594.862260 50.000000 -1594.862274 I set the DM.Tolerance to 1.d-6, so was expecting an energy tolerance of at least 1e-5 eV or better for each calculation. It is therefore puzzling to see there is still noticeable energy difference for such large x and y cell dimensions, especially for y, which I would expect to be flat after 10 Ang, as from the output, the maximum orbital and potential cutoff radii are about 5.1 ~ 5.2 bohr. I used the C.psf and H.psf files downloaded from siesta website, and also tried out my own ones generated using Opium. But all give the same problems. I have also tried the same calculations with Soft-confinement options turned on: PAO.BasisSize SZ PAO.EnergyShift 200 meV PAO.SoftDefault T PAO.SoftInnerRadius 0.9 PAO.SoftPotential 40.0 Ry But the results were even worse: # Cell x dim (Ang) | Total Energy (eV) 10.000000 -1595.419901 15.000000 -1595.445128 20.000000 -1595.445689 25.000000 -1595.445734 30.000000 -1595.445739 35.000000 -1595.445493 40.000000 -1595.445765 45.000000 -1595.445729 50.000000 -1595.445735 60.000000 -1595.445707 70.000000 -1595.445516 # Cell y dim (Ang) | Total Energy (eV) 5.000000 -1595.287697 10.000000 -1595.445739 15.000000 -1595.449755 20.000000 -1595.450680 25.000000 -1595.450917 30.000000 -1595.450999 40.000000 -1595.451061 50.000000 -1595.451042 If anyone have a solution to my problem I would very grateful. This has been bugging me for over a week now, and I need to get thing as basic as this correct or understood to be confident of my calculated results. I have a feeling that it may have something related to the integration grid, as at the moment I am also having problem converging the mesh, even with the egg-box corrections set to: %block GridCellSampling 0.5 0.5 0.5 %endblock GridCellSampling as shown below (in all cases DM.Tolerance = 1.d-6, and it appeared the SCF calculations converged within 20 iterations): # GNR Armchair N=5 pbc Siesta SZ LDA mesh # Planewave equivalent mesh cutoff (Ry) | (eV) | Total Energy (eV) 50.000000 680.279626 -1597.529386 100.000000 1360.559253 -1597.523020 150.000000 2040.838879 -1597.524000 200.000000 2721.118506 -1597.524111 300.000000 4081.677759 -1597.523661 400.000000 5442.237012 -1597.523763 500.000000 6802.796265 -1597.523767 600.000000 8163.355518 -1597.523812 700.000000 9523.914771 -1597.523724 800.000000 10884.474024 -1597.523694 900.000000 12245.033277 -1597.523779 1000.000000 13605.592530 -1597.523773 1100.000000 14966.151783 -1597.523767 1200.000000 16326.711036 -1597.523733 1300.000000 17687.270289 -1597.523751 1400.000000 19047.829542 -1597.523751 1500.000000 20408.388795 -1597.523765 I apologise for the long email, and thank you all for your time and help in advance. Best regards, Lianheng ========================================================== Lianheng Tong Tel: +44 20 7679 3302 London Centre For Nanotechnology Fax: +44 20 7679 0595 University College London [email protected] 17–19 Gordon Street, London WC1H 0AH, U.K. ==========================================================
