Re: [Pw_forum] Convergence of Magnetization in Graphene Monovacancy Supercell
Dear Haricharan Padmanabhan The magnetization asociated with a vacancy is known to converge very slowly. As you will see in the following detailed study: PHYSICAL REVIEW B 85, 245443 (2012) the 6x6 supercell in in fact very small... if you want to get your magnetization converged. 2d systems can have some advantages but also some serious drawbacks due to the very slow convergence of certain quantities related to the bi-dimensionality. This is also why tight-binding is very popular in graphene :-) good luck Cyrille -- Cyrille Barreteau CEA Saclay, IRAMIS, SPEC Bat. 771 91191 Gif sur Yvette Cedex, FRANCE DTU Nanotech Ørsteds Plads, building 345E DK-2800 Kgs. Lyngby, DENMARK +33 1 69 08 29 51 / +33 6 47 53 66 52 (mobile) (Fr) +4545 25 63 12/ +45 28 72 55 18 (mobile) (Dk) email: cyrille.barret...@cea.fr / cyr...@nanotech.dtu.dk Web: http://iramis.cea.fr/Pisp/cyrille.barreteau/ --- De : pw_forum-boun...@pwscf.org [pw_forum-boun...@pwscf.org] de la part de Haricharan Padmanabhan [hari00...@gmail.com] Envoyé : mardi 21 octobre 2014 10:43 À : pw_forum@pwscf.org Objet : [Pw_forum] Convergence of Magnetization in Graphene Monovacancy Supercell Dear Quantum ESPRESSO users, I am attempting to estimate the value of the magnetism in Graphene with a mono-vacancy, using supercells of different sizes. Some background - - One would expect (from literature) the magnetism to converge to around 1.5 bohr magnetons (uB) as the supercell size is increased. - Since vacancies result in localized states at the Fermi level (flat bands, or peaks in the DOS), a dense k-point mesh is usually required to accurately estimate (N.up - N.down), and hence the magnetism. I first obtained convergence with respect to k-point sampling, for a 4x4 supercell (31 atoms + 1 vacancy) K-point meshTotal Energy (Ry) Total magnetization (uB) 16x16 -355.5861.29 20x20 -355.5861.21 24x24 -355.5861.25 32x32 -355.5861.27 36x36 -355.5861.27 A larger 6x6 supercell (71 atoms + 1 vacancy), by conventional wisdom, would require a less dense k-point mesh for convergence. However, even with a dense 32x32 k-point mesh, I get a non-converged value of 0.59 uB for the magnetism. Different calculations with different k-point meshes give me values that oscillate between 0.59 and 1.45 uB, with no apparent pattern. It does not make sense to me to further increase the k-point mesh density. Clearly, the flat bands at the Fermi level are causing trouble depending on whether they've been bumped slightly above or below the Fermi level, due to inadequate k-point sampling in different calculations. How can I fix this problem? Will doing a manual k-point sampling help? A part of the input file - &system ibrav= 4, celldm(1) =27.9, celldm(3) = 1, nat= 71, ntyp= 1, ecutwfc =30.0, ecutrho = 250.0, occupations='smearing', smearing='gaussian', degauss=0.001 nspin = 2, starting_magnetization(1)=0.7 / &electrons diagonalization='cg' mixing_mode = 'plain' mixing_beta = 0.1 conv_thr = 1.0d-6 electron_maxstep = 200 / ATOMIC_SPECIES C 12.011 c_pbe_v1.2.uspp.F.UPF K_POINTS {automatic} 32 32 1 0 0 0 A part of the output file - the Fermi energy is-1.9682 ev total energy =-815.17816366 Ry Harris-Foulkes estimate =-815.17815922 Ry estimated scf accuracy< 0.0077 Ry The total energy is the sum of the following terms: one-electron contribution = -5427.83442348 Ry hartree contribution =2763.25828072 Ry xc contribution =-257.55564014 Ry ewald contribution=2106.95386447 Ry smearing contrib. (-TS) = -0.00024524 Ry total magnetization = 0.59 Bohr mag/cell absolute magnetization= 0.79 Bohr mag/cell Thank you. Haricharan Padmanabhan Indian Institute of Technology Madras ___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum
Re: [Pw_forum] Convergence of Magnetization in Graphene Monovacancy Supercell
On 10/21/2014 10:43 AM, Haricharan Padmanabhan wrote: > Clearly, the flat bands at the Fermi level are causing trouble > depending on whether they've been bumped slightly above or below the > Fermi level, due to inadequate k-point sampling in different > calculations. How can I fix this problem? Will doing a manual k-point > sampling help? Are you increasing the inter-layer space in the larger cell? I cannot tell from the tiny bit of input file you provide. If you do, don't. Too much vacuum space makes the calculation difficult to converge and you may even get electrons in the vacuum. About 6 or 7 Angstroms of vacuum is enough. best regards -- Dr. Lorenzo Paulatto IdR @ IMPMC -- CNRS & Université Paris 6 +33 (0)1 44 275 084 / skype: paulatz http://www.impmc.upmc.fr/~paulatto/ 23-24/4é16 Boîte courrier 115, 4 place Jussieu 75252 Paris Cédex 05 ___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum
[Pw_forum] Convergence of Magnetization in Graphene Monovacancy Supercell
Dear Quantum ESPRESSO users, I am attempting to estimate the value of the magnetism in Graphene with a mono-vacancy, using supercells of different sizes. Some background - - One would expect (from literature) the magnetism to converge to around 1.5 bohr magnetons (uB) as the supercell size is increased. - Since vacancies result in localized states at the Fermi level (flat bands, or peaks in the DOS), a dense k-point mesh is usually required to accurately estimate (N.up - N.down), and hence the magnetism. I first obtained convergence with respect to k-point sampling, for a 4x4 supercell (31 atoms + 1 vacancy) K-point mesh Total Energy (Ry) Total magnetization (uB) 16x16 -355.586 1.29 20x20 -355.586 1.21 24x24 -355.586 1.25 32x32 -355.586 1.27 36x36 -355.586 1.27 A larger 6x6 supercell (71 atoms + 1 vacancy), by conventional wisdom, would require a less dense k-point mesh for convergence. However, even with a dense 32x32 k-point mesh, I get a non-converged value of 0.59 uB for the magnetism. Different calculations with different k-point meshes give me values that oscillate between 0.59 and 1.45 uB, with no apparent pattern. It does not make sense to me to further increase the k-point mesh density. Clearly, the flat bands at the Fermi level are causing trouble depending on whether they've been bumped slightly above or below the Fermi level, due to inadequate k-point sampling in different calculations. How can I fix this problem? Will doing a manual k-point sampling help? A part of the input file - &system ibrav= 4, celldm(1) =27.9, celldm(3) = 1, nat= 71, ntyp= 1, ecutwfc =30.0, ecutrho = 250.0, occupations='smearing', smearing='gaussian', degauss=0.001 nspin = 2, starting_magnetization(1)=0.7 / &electrons diagonalization='cg' mixing_mode = 'plain' mixing_beta = 0.1 conv_thr = 1.0d-6 electron_maxstep = 200 / ATOMIC_SPECIES C 12.011 c_pbe_v1.2.uspp.F.UPF K_POINTS {automatic} 32 32 1 0 0 0 A part of the output file - the Fermi energy is-1.9682 ev total energy =-815.17816366 Ry Harris-Foulkes estimate =-815.17815922 Ry estimated scf accuracy< 0.0077 Ry The total energy is the sum of the following terms: one-electron contribution = -5427.83442348 Ry hartree contribution =2763.25828072 Ry xc contribution =-257.55564014 Ry ewald contribution=2106.95386447 Ry smearing contrib. (-TS) = -0.00024524 Ry total magnetization = 0.59 Bohr mag/cell absolute magnetization= 0.79 Bohr mag/cell Thank you. Haricharan Padmanabhan Indian Institute of Technology Madras ___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum