Dear Xiaoming, I've read your last e-mail, seen the bands for bulk Mo, and ran out of suggestions. Hopefully other members have already worked on the system and can provide more detailed input. Your structures look quite symmetric already.
Best regards, -Salvador ________________________________ From: [email protected] <[email protected]> on behalf of Xiaoming Wang <[email protected]> Sent: Wednesday, May 6, 2015 2:49 PM To: [email protected] Subject: RE: [SIESTA-L] Band structure of MoSe2 Dear Salvador, Thanks so much for your advice and comments! I have compared the band structure of bulk Mo from ELK and Siesta with different pseudos (see the attached figure). In the figure, CA_07 is the pseudo from http://charter.cnf.cornell.edu/psp_files/Mo.psf<https://urldefense.proofpoint.com/v2/url?u=http-3A__charter.cnf.cornell.edu_psp-5Ffiles_Mo.psf&d=AwMFAg&c=JL-fUnQvtjNLb7dA39cQUcqmjBVITE8MbOdX7Lx6ge8&r=n_Y76F1vumEs9EYNHN2gzA5FD9jzyPhrzl3eOzxCHIQ&m=8Gk03Ty8ofFbSZtDYEb4s2VF5EJvY-8kxxy1llS4dmk&s=w_kuCUZlPV9rbDAZtMWyFlmUO_r5Whe6XdxRIpep2Hw&e=>, CA_12 from http://departments.icmab.es/leem/siesta/Databases/Pseudopotentials/Pseudos_LDA_Abinit/Mo_html/Mo.psf<https://urldefense.proofpoint.com/v2/url?u=http-3A__departments.icmab.es_leem_siesta_Databases_Pseudopotentials_Pseudos-5FLDA-5FAbinit_Mo-5Fhtml_Mo.psf&d=AwMFAg&c=JL-fUnQvtjNLb7dA39cQUcqmjBVITE8MbOdX7Lx6ge8&r=n_Y76F1vumEs9EYNHN2gzA5FD9jzyPhrzl3eOzxCHIQ&m=8Gk03Ty8ofFbSZtDYEb4s2VF5EJvY-8kxxy1llS4dmk&s=l13_3_Z8CMODgL-L1VeJuQNyR9DZXSLG0eUg7C085_c&e=>, and I also tried the pseudo by A. Khein and D.C. Allan as you suggested, the result of which is very close to that of CA_12, thus, I don’t show in the Fig. As seen from the Fig, I think both of them are acceptable as for bulk band structure calculations, but there are some discrepancies about the lattice constant. The energy shift of 50 meV was chosen for the calculations, I also tried 10 meV of ES, but nearly no significant change can be found. I didn’t test the bulk Se calculations, as you have done this. The relaxed structure is attached below which is calculated by using CA_07 of Mo pseudo and Se pseudo from Comput. Mater. Sci. , 98 (2015) 372-389, I think it is hexagonal-symmetric, but the K point bands problem still not fixed. And I cannot understand how to re-symmetrize the relaxed structure, can you please explain a little more? By the way, as you mentioned black phosphorus, I also did this kind of calculations before, the problem is not only the gap change with some strain as you said but also there are some split about the bands, it seems that I turned on the SOC during the calculation. Is this also the result of not proper pseudos? Structure of MoSe2: 3.248660546 0.000000000 0.000000000 -1.624330273 2.813422563 0.000000000 0.000000000 0.000000000 13.000000000 3 1 42 0.333333334 0.666666664 0.505632878 2 34 0.666666667 0.333333332 0.632853521 2 34 0.666666667 0.333333332 0.378419220 Best regards, Xiaoming From: [email protected] [mailto:[email protected]] On Behalf Of Salvador Barraza-Lopez Sent: Friday, May 1, 2015 10:08 AM To: [email protected] Subject: RE: [SIESTA-L] Band structure of MoSe2 Dear Xiaoming, Okay, I see. I have some further comments. You set a relaxation with cell parameters evolving: That setting could be responsible for shifting valleys away from the K-point (this similar to the effect of K-point shifting induced by strain; this has been discussed extensively on other materials with similar structural symmetries and where the K-point is special, such as graphene; this is a known effect). So you want to re-symmetrize the relaxed structure, especially if the lattice vectors change. This may fix the issue near K. In addition, there is no Mo pseudo on the article you mention, only Se. WSe2 is shown as Fig 15 there. Note that there are some discrepancies with VASP results on that Figure, and on the bands for W in Figure 10 there as well, but they are not nearly as dramatic as the ones you show for MoSe2. Once your MoSe2 structures are strictly hexagonal-symmetric and fine, may I suggest that you compare lattice parameters and band structures for bulk Mo with SIESTA and with another code (ELK or Wien2K -all electron codes- could work for this purpose). You may want to do the same for Se (we did this for Se on the paper you mention, and the datasets are attached as a PDF file here: http://www.sciencedirect.com/science/article/pii/S2352340914000353<https://urldefense.proofpoint.com/v2/url?u=http-3A__www.sciencedirect.com_science_article_pii_S2352340914000353&d=AwMFAg&c=JL-fUnQvtjNLb7dA39cQUcqmjBVITE8MbOdX7Lx6ge8&r=n_Y76F1vumEs9EYNHN2gzA5FD9jzyPhrzl3eOzxCHIQ&m=8Gk03Ty8ofFbSZtDYEb4s2VF5EJvY-8kxxy1llS4dmk&s=Ij4NECQWU-Trdrz4aRYSQCDT_bqt78DrJhnrJkdxwy4&e=> ; we did not work with Mo though). Here, you have to recall as well that band structures evolve dramatically with "strain" in 2D materials; so if your lattice constant is 1% off with respect to ELK or Wien2K (ie, your optimal parameters are "strained" as seen from results using other codes), you may be seeing gap closings as large as 5% with respect to the results of that code as well as other effects in the bands (this is another well-known drastic effect of lattice parameters in 2D; this is especially seen on black phosphorus). So to help with Mo, another pseudo you want to try out is the Troullier-Martins one generated by A. Khein and D.C. Allan: http://www.abinit.org/downloads/psp-links/psp-links/lda_tm<https://urldefense.proofpoint.com/v2/url?u=http-3A__www.abinit.org_downloads_psp-2Dlinks_psp-2Dlinks_lda-5Ftm&d=AwMFAg&c=JL-fUnQvtjNLb7dA39cQUcqmjBVITE8MbOdX7Lx6ge8&r=n_Y76F1vumEs9EYNHN2gzA5FD9jzyPhrzl3eOzxCHIQ&m=8Gk03Ty8ofFbSZtDYEb4s2VF5EJvY-8kxxy1llS4dmk&s=8zhmqZUEg2WPXmeSEEg_Rkmb4K1O0sO_HeDAbeiAMqI&e=> For example, the radii in your LDA Mo INP file would have to be: 2.8930462, 3.1973167, and 2.1974630, as found here: ftp://ftp.abinit.org/pub/abinitio/Psps/LDA_TM.psps/42/42mo.pspnc<https://urldefense.proofpoint.com/v2/url?u=ftp-3A__ftp.abinit.org_pub_abinitio_Psps_LDA-5FTM.psps_42_42mo.pspnc&d=AwMFAg&c=JL-fUnQvtjNLb7dA39cQUcqmjBVITE8MbOdX7Lx6ge8&r=n_Y76F1vumEs9EYNHN2gzA5FD9jzyPhrzl3eOzxCHIQ&m=8Gk03Ty8ofFbSZtDYEb4s2VF5EJvY-8kxxy1llS4dmk&s=C2fXoFGHuY6Nv-Xcaanc1UYKkOYfem8Fh2EZJfvIh-s&e=> That procedure worked for Pt, but you want to see if it does the job for you. For Mo LDA, these radii currently are 2.75000, 2.89000, and 2.49000, as it can be seen on the INP file that you've used at http://departments.icmab.es/leem/siesta/Databases/Pseudopotentials/Pseudos_LDA_Abinit/Mo_html/Mo.html<https://urldefense.proofpoint.com/v2/url?u=http-3A__departments.icmab.es_leem_siesta_Databases_Pseudopotentials_Pseudos-5FLDA-5FAbinit_Mo-5Fhtml_Mo.html&d=AwMFAg&c=JL-fUnQvtjNLb7dA39cQUcqmjBVITE8MbOdX7Lx6ge8&r=n_Y76F1vumEs9EYNHN2gzA5FD9jzyPhrzl3eOzxCHIQ&m=8Gk03Ty8ofFbSZtDYEb4s2VF5EJvY-8kxxy1llS4dmk&s=MXVJyukpicR0AUlYeUOMd4Q52ldd7IqzTmQZ_c0-Eco&e=> pg -- file generated from Mo ps file tm2 Mo ca 0.000 0.000 0.000 0.000 0.000 0.000 8 4 5 0 1.000 0.000 #5s 5 1 0.000 0.000 #5p 4 2 5.000 0.000 #4d 4 3 0.000 0.000 #4f 2.75000 2.89000 2.49000 2.49000 0.00000 0.00000 #23456789012345678901234567890123456789012345678901234567890 Ruler you can leave the last channel to be 2.49, that f-channel will not change things for you.) For layered materials, we have also noticed some discrepancies between bulk and single-layer results; here is where basis size could also play a role, and hence my previous question. So once again, I would ensure the structure is properly symmetric after relaxation, and I would try to reproduce the properties of bulk Mo first, and then use that pseudo for MoSe2: There is some additional work ahead for you. If the symmetries are correct, then you've got to work with the Mo pseudo; there will probably be no way around it. Hopefully this e-mail sets a solid pathway for you to look at. I will be happy to learn on your progress. Best regards, -Salvador ________________________________ From: [email protected]<mailto:[email protected]> <[email protected]<mailto:[email protected]>> on behalf of Xiaoming Wang <[email protected]<mailto:[email protected]>> Sent: Thursday, April 30, 2015 9:25 PM To: [email protected]<mailto:[email protected]> Subject: RE: [SIESTA-L] Band structure of MoSe2 Hi Salvador, Thanks for your reply. Changing the shifts to 0.0 0.0 0.0 actually can’t solve the problem. And I have also checked the basis sizes of SZ, SZP, DZ but not beyond DZP, the ‘non-differential’ point still exists. By the way, I have rechecked the band structure, the ‘non-differential’ points can actually also be observed at M(0.0, 0.5, 0.0) k-point, but not as obvious as that of K(0.333, 0.333, 0.0) point. I havn’t tried the bulk band structure yet. Best regards, Xiaoming From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Salvador Barraza-Lopez Sent: Thursday, April 30, 2015 9:46 PM To: [email protected]<mailto:[email protected]> Subject: RE: [SIESTA-L] Band structure of MoSe2 The only thing that is evident to me at this moment is that you are avoiding the K-points in your k-point sampling; as indicated by Marco Verisimo a number of years ago your number of k-points must be a multiple of three, which it is, but the shift by 0.5, 0.5 is driving the sampling away from the K-points and avoiding them... Hopefully bringing the shifts to 0.0 0.0 0.0 solves the "differential," or the missing of the K-point on the conduction band. I would work on the Mo pseudo a little more; and have you checked the effect of the basis size as well? Do your pseudos give you a reasonable bulk band structure? Best regards, -Salvador ________________________________ From: [email protected]<mailto:[email protected]> <[email protected]<mailto:[email protected]>> on behalf of Xiaoming Wang <[email protected]<mailto:[email protected]>> Sent: Thursday, April 30, 2015 7:23 PM To: [email protected]<mailto:[email protected]> Subject: [SIESTA-L] Band structure of MoSe2 Dear Siesta users, Recently, I’m trying to calculate the band structure of monolayer MoSe2. However, when I plotting the bands, the band curves at K point are not at maximum or minimum, which should be. And there seems a non-differential point at K along the bands. The band structure of MoSe2 is attached, the non-differential point is more obvious at the lowest two bands shown in the Fig. Both LDA and PBE functionals give this strange phenomenon. I have tried to tune many parameters, but failed to get good result. Can anyone help me with the problem? By the way, I have tried pseudopotentials from Siesta website and from http://charter.cnf.cornell.edu/<https://urldefense.proofpoint.com/v2/url?u=http-3A__charter.cnf.cornell.edu_&d=AwMFAg&c=JL-fUnQvtjNLb7dA39cQUcqmjBVITE8MbOdX7Lx6ge8&r=n_Y76F1vumEs9EYNHN2gzA5FD9jzyPhrzl3eOzxCHIQ&m=NaDRSvcZImjiuqSJSGFoSEOGLfumFBd6_fRZa0Hs0Bs&s=UQq2ZNZfUJL5kmP2F3B0GlcPmhnT4dJuFcGi_NHsl2Q&e=>, and also used the recently published pps of Comput. Mater. Sci. , 98 (2015) 372-389. I also changed the energshift, meshcutoff, and k point mesh. But none of them can resolve the problem. Below are the input files for relaxation and band structure calculations. Any comment or advice are highly appreciated. fdf for relaxation: # General System Descriptors SystemName mose2 SystemLabel mose2 NumberOfAtoms 3 NumberOfSpecies 2 %block ChemicalSpeciesLabel 1 42 Mo 2 34 Se %endblock ChemicalSpeciesLabel PAO.EnergyShift 50 meV PAO.BasisSize DZP # Structure and K-sampling LatticeConstant 1.00 Ang %block LatticeParameters 3.25 3.25 20.00 90. 90. 120. %endblock LatticeParameters AtomicCoordinatesFormat Fractional AtomicCoorFormatOut Ang %block AtomicCoordinatesAndAtomicSpecies 0.333333333 0.666666670 0.505669950 1 0.666666667 0.333333330 0.625537344 2 0.666666667 0.333333330 0.385811946 2 %endblock AtomicCoordinatesAndAtomicSpecies %block kgrid_Monkhorst_Pack 12 0 0 0.5 0 12 0 0.5 0 0 1 0.0 %endblock kgrid_Monkhorst_Pack # DFT XC.functional LDA XC.authors CA SpinPolarized false MaxSCFIterations 200 DM.MixingWeight 0.25 DM.NumberPulay 3 DM.Tolerance 1.d-5 MeshCutoff 300 Ry SolutionMethod diagon ElectronicTemperature 300.0 K # MD and Relaxations MD.TypeOfRun CG MD.VariableCell T MD.MaxForceTol 0.01 eV/Ang MD.MaxStressTol 0.1 GPa MD.NumCGsteps 200 MD.MaxCGDispl 0.1 Ang MD.RelaxCellOnly F %block GeometryConstraints stress 3 4 5 6 %endblock GeometryConstraints # Output options WriteCoorInitial true WriteCoorStep true WriteForces true WriteKpoints false WriteEigenvalues false WriteKbands false WriteBands false WriteMullikenPop 0 WriteWaveFunction false WriteCoorXmol true WriteCoorCerius false WriteMDCoorXmol true WriteMDhistory true WriteMDXmol true WriteDM true fdf for band structure: # General System Descriptors SystemName mose2 SystemLabel mose2 NumberOfAtoms 3 NumberOfSpecies 2 %block ChemicalSpeciesLabel 1 42 Mo 2 34 Se %endblock ChemicalSpeciesLabel PAO.BasisSize DZP PAO.EnergyShift 50 meV # Structure and K-sampling LatticeConstant 1.00 Ang %block LatticeParameters 3.16 3.16 13.00 90. 90. 120. %endblock LatticeParameters AtomicCoordinatesFormat Fractional AtomicCoorFormatOut Ang %block AtomicCoordinatesAndAtomicSpecies 0.333333333 0.666666670 0.505669950 1 0.666666667 0.333333330 0.625537344 2 0.666666667 0.333333330 0.385811946 2 %endblock AtomicCoordinatesAndAtomicSpecies %block kgrid_Monkhorst_Pack 24 0 0 0.5 0 24 0 0.5 0 0 1 0.0 %endblock kgrid_Monkhorst_Pack # DFT XC.functional LDA XC.authors CA SpinPolarized false MaxSCFIterations 200 DM.MixingWeight 0.25 DM.NumberPulay 3 DM.Tolerance 1.d-5 MeshCutoff 300 Ry SolutionMethod diagon ElectronicTemperature 300.0 K # Output options WriteCoorInitial true WriteCoorStep true WriteForces true WriteKpoints false WriteEigenvalues false WriteKbands false WriteBands false WriteMullikenPop 0 WriteWaveFunction false WriteCoorXmol true WriteCoorCerius false WriteMDCoorXmol true WriteMDhistory true WriteMDXmol true WriteDM true UseStructFile T BandLinesScale ReciprocalLatticeVectors %block BandLines 1 0.0000 0.0000 0.0000 \Gamma 70 0.0000 0.5000 0.0000 M 40 0.3333 0.3333 0.0000 K 80 0.0000 0.0000 0.0000 \Gamma %endblock BandLines Best regards, Xiaoming Wang IAMDN, Rutgers
