First-principles Calculations

First-principles Calculations

  • Federico Brivio added an answer:
    11
    How large supercell should I build in order to calculate phonon?

    Hi, Folks!

    I am going to use finite displacement (sometimes called finite difference if I am right) to calculate phonon of my crystal. As I read from some articles and online forums, we should build a supercell first, then move one atom each atom by a given distance. I have two questions concerning supercell.

    1)How could I know how large supercell (2*2*2 or 3*3*3 or even larger) I need to  build to satisfy the phonon calculation?

    2)What's more, should I also increase k-points correspondingly? For example, if I use 4*3*3 for unit cell, then I need to increase to 32*24*24 in case of 2*2*2 supercell.

    Federico Brivio

    I meant VASP. If you change software you need to reoptimize. About the Gamma vs number of atoms, depends on the structure. If you have 80 atoms, but a small dimensionality in one direction, in that direction you might be interested in the dispersion.

  • Maximo Ramirez asked a question:
    New
    How can I get accuracy optimize structure?

    Hello,

    I'm trying to get accuracy optimize structure, but the lattice constant is not consistent with experimental data.

    For example:

    I'm working with spinel LiMn2O4 (space group fd-3m:2), like show in attached cif file, with 56 atoms (8 atoms of Li, 16 atoms of Mn and 32 atoms of O).

    The lattice constant in this cif file is 8.2399 A, this value is consistent with value obtain by other researches.

    I'm optimizing the structure in VASP, but the lattice constant value that I got  8.0625  (not consistent).

    I'm using PAW-PBE pseudopotencials and 8x8x8 k-points grid. Bellow INCAR that I used.

    How can I get more accuracy lattice and structure?

    INCAR

    ******************************

        SYSTEM = LiMn2O4
        ENCUT = 520 #1.3 x ENMAX
        PREC = Normal
        LREAL = Auto

        EDIFF = 1E-5
        EDIFFG = -0.01
        NSW = 10


        ISMEAR = 2; SIGMA = 0.2;
        IBRION = 2
        ISIF = 3

        NCORE= 4

    #save disc space
        LWAVE = .FALSE.
        LCHARG = .FALSE.

    ******************************

  • Vaibhav Kaware added an answer:
    16
    How to define a crystal structure in quantum-espresso (pwscf)?

    In the tutorial of quantum-espresso (pwscf), the tutor demonstrated how to set up a unit cell using the parameters list below.

    ibrav=2,

    nat=1,

    ntyp=1

    Only 1 atom is enough to define this face-centered cubic. As far as I understand from this parameter, the software will know where to place 14 atoms in one cubic (8 corners and 6 face centers).

    However, I was confused by the second example. The tutor only used 2 atoms to define a diamond structure (Fig 1).

    ibrav=2,

    nat=2,

    ntyp=1

    Atomic_position

    Si 0.0 0.0 0.0

    Si 0.25 0.25 0.25

    If the first Si was considered as the atoms whom should be placed in the 8 corners and 6 face centers in one cubic, how does the software know where to place the second Si? How does the software distinguish the diamond structure from fluorite structure (Fig2)?

    + 1 more attachment

    Vaibhav Kaware

    Dear Sinkander,

      I typically use these options. Check if this helps.

    cif2cell -f nis2-1991.cif -p quantum-espresso -o nis2-1991.in --cartesian --coordinate-tolerance=0.01 --setup-all

    Please also consider the small but finite possibility that  " case.cif "  may not comply with the cif format and/or is erroneous.

  • Muhammad Ali Askari added an answer:
    14
    Non-equilibrium Green's function based calculations?

    How can we perform Non-equilibrium Green's function based calculations? please recommend some simulation package for Non-equilibrium Green's function based calculations with its procedure.

    Is it possible to perform Non-equilibrium Green's function based calculations by vasp?

    Muhammad Ali Askari

    Thank you Dr. Hamidreza Simchi for your kind suggestions.

  • Muhammad Ali Askari added an answer:
    5
    How can we calculate the absorption spectra by vasp?


    How can we calculate the absorption spectra by vasp? Please recommend some literature for better understanding of absorption spectra and its relation with band gap/electron transition.

    Muhammad Ali Askari

    Dear David, 

    Could you please guide me how to script the code to calculate optical properties?

  • Muhammad Ali Askari added an answer:
    18
    How can I select the correct value for NSW in VASP?

    I'm studying effect of doping LiM2O4 (56 atoms) spinel, I relaxed the structure without doped and took me 7 ionic steps to get this message

    reached required accuracy - stopping structural energy minimisation

    But now I'm doping this spinel with Cr, Fe, Co, etc. at 6.25% (e.g. LiCr0.125Mn1.875O4) and I need structural optimization. I set NSW=50 and this took a lot of time but I didn't get the message

    reached required accuracy - stopping structural energy minimisation

    After ionic step 50 the calculation stopped.

    I talk about vasp, but I believe ionic step is related to most ab initio codes.

    Is it necessary in VASP to run the system upto (NSW value) it says reached required accuracy - stopping structural energy minimisation?

    Muhammad Ali Askari

    Thank you for making it clear @ Shaira Narido.

  • Vladimir S. Myasnichenko added an answer:
    2
    Tight Binding Calculations?

    How can we perform tight binding calculations? Any simulation package?

  • Andrew O'Hara added an answer:
    1
    Curie Temperature Calculation by DFT/MD calculations (vasp)?

    How can we calculate curie temperature through vasp calculations? 

    Anyone can suggest how to initiate Monte Carlo simulations for curie temperature calculations for comparison?

    Andrew O'Hara

    One method for calculating the Curie temperature of a ferromagnetic using density functional theory is to construct model spin Hamiltonian to treat in terms of a mean-field type approach (in this case the Curie Weiss model). The model will be parameterized by J, the exchange constant between magnetic atoms (with different J's for 1st nearest neighbors, 2nd nearest neighbors, etc).  Essentially in this step, you would need to do DFT calculations for a variety of magnetic configurations and then use the energies to extract the values of J.  An example of this was done in the following paper for EuO and Gd-doped EuO:
    J. Lee, N. Sai, and A. A. Demkov "Spin-polarized two-dimensional electron gas through electrostatic doping in LaAlO3/EuO heterostructures," Phys. Rev. B 82, 235305 (2010)

    A more detailed description of how to construct the effective Hamiltonian and use DFT to extract the parameters can be found in Jaekwang Lee's dissertation in Chapter 5 (in particular sections 5.4.6 through 5.4.8 (p58-p64)).  The dissertation is available from the University of Texas here:
    https://repositories.lib.utexas.edu/handle/2152/ETD-UT-2010-08-2012

  • Bharat Medasani added an answer:
    4
    How can I prevent convergence issues in nudged elastic band calculations?

    I am a novice in nudged elastic band calculations and I am attempting to model oxygen diffusion at oxide/metal interface. For this I am using NEB method as implemented in VASP. However the calculations are not converging. Upon inspecting the output, I noticed a message "The distance between some ions is very small." I manually adjusted the positions of those atoms that are very close to a reasonable degree but the message is still displayed.

    1) Is there any script that I can use which automatically adjust the positions of  more than 2 atoms based on some criteria?

    2) Also I am following the advice in vasp neb example and starting with 1 image. However the interpolated structure has couple of atoms very close to each other. Instead of attempting (1), can I generate lot of interpolating structures and use only those that are close to the end points. For example, can I generate 8 images with the interpolating script supplied and use only the two images close to the end points and discard the rest? By repeating this procedure with the optimized images, I am hoping to get to the transition state eventually.

    Bharat Medasani

    Thanks guys for the replies. 

    @Vaibhav, thanks for the link.  I will give the script a try. I initially started with 4 or 5 images, but again I faced the convergence issue with not even a single electronic convergence. Also my supercells are big (>250 atoms due to interface), hence I was spending so much time playing with VASP parameters to get convergence. So I went down to two and one images because I didn't understand the source of trouble initiially.

    @Peter, Yes, I read about CI-NEB method and have it already installed. I was using it initially, but now  planning to use it once I have convergence with NEB method. 

    By the way, the method I suggested as option 2 is working. I generated 16 images and discarded all except the 2 images that are close to the end points. The calculations haven't finished yet, but I am not noticing major problems with convergence.  

    Another question I have is on how to refine the NEB calculations.  I see two ways to proceed. Once the images converge to the minimum energy path (MEP), I can generate new images by interpolating the optimized images again. 1) After generating the new images, shall I keep the prior images converged in earlier run as end points and converge the new images as suggested in vasp manual? 2) Or do I incorporte the old and new images into a single NEB calculation with tthe original end points? Do the converged images move on MEP due to the inclusion of additional images? I understand option (1) is easy on resources and also may be efficient because I don't have to waste cpu time on already optimized images.

    Thanks again. 

  • Drahomir Chovan added an answer:
    1
    How can I get the crystal structure at ambient temperature by first-principle calculation?

    Usually, we can easily get the structure (0K) by DFT. But I want to gain the structure at 300K. The only difference between these structures is a slight displacement of atoms, whose positions become essentially equal after ionic relaxation. 

       When we run MD in VASP, we only use nvt (no npt). I can use nvt to get structure? Whether we can neglect the pressure influence from nvt?

      Or, there are some better method to do it?

    Drahomir Chovan

    ...The only difference between these structures is a slight displacement of atoms, whose positions become essentially equal after ionic relaxation.... what does this mean? BTW, your approach is not correct at all. you dont want to keep volume constant and have build-up pressure.

  • Thanayut Kaewmaraya added an answer:
    5
    Can anyone help with the following issues about hse06 hybrid functional calculations using vasp 5.3.5?

    I am not asking here how to set up the hybrid functional calculations (with hse06) in vasp, which has been outlined in many places for example the vasp wiki. I encountered the problem that exactly same set up working well in vasp 5.2.12 did not work for vasp 5.3.5.

    I started with a PBE functional calculation and saved the WAVECAR. Then I perform the HSE06 calculation reading the wavefunction and using Damped keyword to do the digonalization. While the vasp 5.2.12 and 5.3.5 presented the same energy in the PBE calculation, the same setup converged quickly in 5.2.12 within roughly 15 steps but diverged in 5.3.5 after three or four steps. Indeed, the first three steps in 5.3.5 did make sense with a negtive energy close to the converged one in 5.2.12. But for some reasion, the iteration just diverged and large positive energy was yielded afterwards. I read the mannual and did not find significant change in the INCAR for vasp 5.2.12 and 5.3.5.

    I have tried to google the similar topic and found no one mentioned the same problem as I had. Has anybody here noticed my problem and can you suggest some hints? Thanks in advance

    A portion of my INCAR for the HSE06 calculation.

    ISTART=1
    LHFCALC=.TURE.
    GGA=PE
    HFSCREEN = 0.2

    TIME = 0.4
    AEXX = 0.25
    AGGAX = 0.75
    AGGAC = 1.0
    ALDAC = 1.0
    ALGO = Damped
    PRECFOCK = Fast !--> used PRECFOCK = Normal for high quality calculations

    Thanayut Kaewmaraya

    Dear Bin Jiang

    I am also facing the same problem by VASP 5.3.x. In fact, there are some important parameters (e. g. ALGO and TIME) for HSE06 calculations depending on the system you are investigating, i. e. insulators or semiconductors.  For my case (a semiconducting nanostructure), the total energy in OSZICAR is diverged if choosing the algorithm ALGO = Damp. After switching to ALGO = All with TIME = 0.3, the total energy is normally converged to the desired accuracy. In addition, the number of bands, NBANDS, required for HSE06 should be significantly greater than those used in the conventional PBE.

    I hope this may help you and others.

    Best regards,

    Thanayut

  • Jianqiu Huang added an answer:
    8
    How does vacancy concentration effect vacancy migration barrier?

    Using VASP I calculated the activation barrier of vacancy diffusion on the surface of ZnO for a 2x2 3 layered cell, where i had removed 1 oxygen atom out of 4 on the surface, and it came out to be 1.35eV and then I calculated the migration barrier for a 3x3 3 layered cell, where I had removed 1 out of the 9 oxygen atoms on the surface, the barrier came out to be 1.67 eV.

    The thing which is bugging me is, why is the barrier higher at a lower vacancy concentration. Is there an explanation for this? I was wondering if i should use a similar sized cell for both the calculations ( a 6x6 cell)? The only thing that is different in the two calculations are the KPOINTS. For the 2x2 cell I have used a 2x3x1 mesh and for the 3x3 cell I have used a 1x2x1 mesh.

    Jianqiu Huang

    Hi Salman, 

    I will be calculating the same property in SiO2 using vasp. Since I am new in this field, can you please detail me how do you calculate vacancy migration barrier?

    Thanks very much

  • Salman Ahmad Khan added an answer:
    3
    Does increasing the number of images improve convergence when doing a NEB calculation using VASP? Any other pointers for improving convergence?

    I want to calculate the path of diffusion of surface vacancies in ZnO. I was able to get converged results for a 2x2 3 layered cell using 5 images, but the same calculation didn't converge with 3 images. Now I want to do the same calculations, but with a bigger cell ( 3 layered 3x3 cell). I have tried using 5 and 7 images, but the calculations just won't converge. I preconverged all the images using IBRION = 3 and I converged the end points using IBRION =1 and I am using IBRION =1 for the NEB run. I am attaching the input files.

    Any pointers?

    Salman Ahmad Khan

    Thanks for the help
    I was able to get converged results for 7 images by freezing the bottom layer, though it took 941 steps. I did not use the converged path though. I gave 9 images a shot too, but i was surprised that the calculations did not converge. I will try doing the calculations using the converged path now and next I will try the climbing image method. Thanks again

  • Hamidreza Simchi added an answer:
    4
    Is Rashba SOI or first principle calculation more suitable for a surface system?

    Dear all

    In one of my unpublished paper, I used second order perturbation theory to calculate the magnetic anisotropy of a Ni surface. The unperturbated wavefunction was calculated by colinear DFT. The detail formula  can be found in Phys. Rev. Lett. 99, 177207.I think as long as the perturbation converged, the Rashba and Dresselhaus term has been included in the calculation. Because the information of symmetry breaking is contained in the charge density (or orbital occupancy) of colinear DFT. However, somebody told me that I should additionally include a Rashba term to make my calculation suitable for a surface system.  So I am confused. Who is correct?

    Hamidreza Simchi

    Hi!

    We know that, the electric field of core is seen as magnetic filed in rest frame of electron as   B= - Gama ( beta x E), where beta=V/C. It is internal SOC and is written as:

    HI= - meyouB. ( B.S) where S is spin. 

    When electron jumps between two neighbor atoms, and there is an external electric field, Rashba term appears as:

    HR= - lambdaR . (sigma . (E x p)) , where sigma is Pauli's matrices and p is momentum. 

    When one apply an approximation on Dirac equation, she/he can show that these terms will be appeared in Hamiltonian.

    Now, if your DFT code, uses complete Hamiltonian i.e., 

    H=H0 + HI + HR

    and assumes HI and HR are perturbation terms it is not necessary you consider them again. But if not, you should add them.

    Best Regards 

  • Sara Kadkhodaei added an answer:
    10
    Does anyone know how to calculate free energy using quasi-harmonic Debye model using PHONOPY?
    For some unstable states, one should expect negative phonon frequencies in the harmonic approximations. If you have negative phonon frequency then you can't get the free energy using that. For that, you have to use quasi-harmonic Debye model.
    Now, in my case, I want to calculate those using PHONOPY program package. Can anybody help me?
    Sara Kadkhodaei

    The following paper resolve the problem:

    http://journals.aps.org/prb/abstract/10.1103/PhysRevB.80.104116

  • Yin Li added an answer:
    9
    How to place molecules whose atomic position are written in Cartesian coordinate system into unit cell using material studio software?

    I have a structural information (in Cartesian coordinate system) of a small molecule. According to the literature, one unit cell contains three molecules. I have set the parameters of the unit cell such as space symmetry, a and c. How can I set the atomic positions of my molecule in the next step? As I know, those positions should be written in fractional coordinate system other than Cartersian coordinate system. What's more, how can I set the orientation of 3 molecules in one unit cell? Is there any rule to follow?

    Yin Li

    Thanks Stephen! It works! 

  • V. V. Bannikov added an answer:
    2
    What are the effects of high electron density at the fermi level regarding the oxygen reduction reaction?

    I recently synthesized a doped cobalt oxide and high electron density (dxy) is observed right at the fermi level through first principle calculations. Does anyone know what are the effects of high electron density at fermi level? 

    V. V. Bannikov

    It the first-principles calculations are ordinary DFT calculations they may highly overestimate the DOS at Fermi level for the compounds with open-shell transition metals.

  • Sayyed Faramarz Tayyari added an answer:
    5
    How to properly apply Gaussian G2 method?
    I would like to ask what is the proper way to use this Gaussian composite method for high-accurate energy calculations? Do I need to pre-optimize my molecule using the same functional and basis set (in this case: MP2/6-31G*) as its required for starting geometry in G2 method? Or maybe this pre-optimization step is done automatically within G2 method and there is no difference what functional/basis set was used to prepare initial structure? Moreover, should I use OPT keyword or just G2 alone? I noticed that in case of correcting energies of transition state structures, the OPT(TS) keyword was needed, otherwise the molecule was optimized either to reactants or products.
    Sayyed Faramarz Tayyari

    Simply type G2 or G3 then everything goes automatically. However, you cal also apply 

    solvent effects by adding, for example, SCRF=(PCM,solvent=...).

  • Jian Hui Lan added an answer:
    8
    How do you calculate the energy of a gas phase oxygen molecule in its triplet ground state?

    I would like to calculate the oxygen vacancy formation.

    Eform,vac = Edefective - Eperfect - 1/2 EO2

    I read that the energy of oxygen vacancy was referenced to the oxygen gas energy in its triplet ground state. I would like to know how to calculate the triplet ground state O2 energy?

    I heard that the O pseudopotential should be test by binding energy and bond length of O2 molecule? Would you please give me some suggestions or references?

    Jian Hui Lan

    Dear Pro. Fiorentini,

    I have to say I learned much from your latest reply. Thanks!

  • Ahmad Bahmani added an answer:
    7
    How can I find out the atomic position of a crystal with known space group(using http://www.cryst.ehu.es/ website)?

    In the following link, the atomic position of different space groups has been shown.

    http://www.cryst.ehu.es/cgi-bin/cryst/programs/nph-wp-list

    http://www.cryst.ehu.es/cgi-bin/cryst/programs/nph-wp-list?gnum=71

    For a first principle calculation, I need to know the atomic position of a compound. I found the crystal space group (e.g. Immm (Pearson number:71)). But for adding atoms in materials studio, I don't know how I should find the atomic position.

    I really appreciate if you help me to find out how to use this website. What should be replaced by x, y, z?

    Best wishes,

    Ahmad Bahmani

    Dear Sir, 

    Thank you so much, as you said the crystal of Mg7Zn3 is same as Mg51Zn20.

    For the case of Ca2Mg6Zn3 I found the following article. 

    DOI: 10.1017/S1431927602010413

    But the simulation is so lasting long. But the simulation takes so long. 

    Best regards

  • Andrij Vasylenko added an answer:
    6
    How to find Cartesian atomic coordinates of Silicon in a unit cell on a surface?
    How can one find these or reciprocal coordinates for a couple of layers in a slab?
    Andrij Vasylenko

    I would build a unit cell according to ME Fleet (1981) Acta Crystallographica B, 37, 917 :

    Space Group Fd-3m (No. 227)
    a=8.3941 Å
    a=90.00
    Z=8

    Atomic Positional Parameters
    Fe1 8a 0.1250 0.1250 0.1250
    Fe2 16d 0.5000 0.5000 0.5000
    O 32e 0.2549 0.2549 0.2549

    With a software like Avogadro it is possible to fill the unit cell, multiply it and to build a slab.

  • Giuseppe Pintaude added an answer:
    11
    How could hardness of metals or alloys be quantified or measured theoretically?

    From the first-principles calculation or basics of dislocation movements, density or anything else aiding to introduce hardness of metals and alloys.

    Giuseppe Pintaude

    Very interesting paper Li! Congrats.

  • Larraga-Gutierrez Manuel Jose added an answer:
    9
    What is the good software for nonlinear curve fitting rather than Excel and Origin?
    What is the good software for nonlinear curve fitting rather than Excel and Origin?
    Larraga-Gutierrez Manuel Jose

    I recommend gnuplot, free software with a lot of features. Easy to use.

  • Zeyad Adnan Saleh added an answer:
    3
    What type of analysis we can do to except band gap measurement using UV-visible spectroscopy?

    I would like to do first principle studies(using DFT) for this technique to compare experimental and theoretical value.

    Thank you.

    Zeyad Adnan Saleh

    yes , you can do that by taking the absorption spectra and the electrical properties of thin films for the materials which are under investigation. we do do that when we studied the effect of doping  in aromatic system.

  • Waheed Adeniyi Adeagbo added an answer:
    6
    How can I calculate the defect formation energy by DFT?

    The paper "first principle calculations for point defects in solids, Rev. Mod. Phys. 86 (1)" gives the defect formation energy formula:

    Ef[D(i,q)] = Etotal[D(i,q)] – Etotal[defect_free] + Sum[n(i)*μ(i)] + q*E(Fermi) + ∆q

    μ(i) is the corresponding chemical potential of the ith point defect

    E(Fermi) is the charge potential

    ∆q is "a correction term that accounts for finite k-point sampling in the case of shallow impurities, or for elastic and/or electrostatic interactions between supercells."

    I know how to get the total energy of perfect and defect structures. but with respect to the latter three items μ(i), E(Fermi), ∆q how to calculate?

    Thanks in advance!

    Waheed Adeniyi Adeagbo

    By following P hys. Status Solidi RRL, 1-5 (2014) / DOI 10.1002/pssr.201409032

    and

    Modelling Simul. Mater. Sci. Eng. 17 (2009) 084002 (14pp)


    E_formation (D,q) = E (D,q) - E(host defect free) + Sum[n(i)*µ(i)] +

                                                q( Ev + delta E_F )


    E_formation (D,q) = formation energy of point defect D with supercell charge q


    E (D,q) = total energy of the point defect D with supercell charge q

    E(host defect free) = E(host defect free) = total energy of host supercell
    containing equivalent number of atoms as that of the
    defect supercell

    Sum[n(i)*µ(i)] = The growth conditions for the crystal are
    incorporated through the chemical potentials µ(i)
    of the elemental species with the number n(i) of atoms added
    ( n(i) negative) or removed ( n(i) positive) from the pure host


    Ev + delta E_F: The Fermi level E_F is referenced with respect
    to the valence band maximum Ev and is related to Ev as
    Ev+ delta E_F (from pure host caluculations)

    |<----------------------->|
    Ev            E_F          Ec

    <------Eg---------------->
    0   (if EvBMmax)    Ec

    Ev = valence band maximum
    Ec = conduction band minimum
    Eg = band gap

    .i.e. Ev < E_F < Ec
    if we set Ev to zero, then E_F is between 0 and Eg (as varying parameter.)

    Now what is Ev?
    In principle, Ev can be obtained from total energy difference between neutral calculation E(n) and one charge calculation
    E(n-1) FOR A VERY LARGE SUPERCELL

    Ev = E(N) - E(N-1) for N---->large


    Or simply do a test neutral calculation with a unit cell.
    Then subtract one electron SLOWLY.
    i.e. from delta(q) = 0e to 1e removed in the interval delta(q) such as of 0.01.

    Then plot the total energy

    E(delta(q)=0)- E(delta(q)) vs log10(1/delta(q))


    A curve with saturation value will be obtained. This saturated value should gives Ev (It is equivalent to the energy needed
    to add or remove a charge to alter the Fermi energy in the charged calculations)

    In VASP e.g. if one plots the DOS, the Fermi Energy obtained from OUTCAR should be equal to Ev (IF the fermi level
    lies exactly at the VBM, OTHERWISE one needs to shift the Fermi level to the VBM. The amount of this shift is subtracted from Fermi energy obtained in OUTCAR file.

    One could compare all these numbers in the 3 methods described.

    e.g. Consider ZnO with 96 atoms total, remove One Oxygen atom and put charge q
    Then we have ZnO:Vo:q

    E_formation (ZnO:Vo:q) = E_tot_supercell(ZnO:Vo:q) - [ E_tot_supercell (ZnO) - 1µ(one O) ] +q ( Ev + delta E_F)

    i.e
                      E(95 atoms) -  [E(96 atoms of  pure)  minus E(one oxygen atom) ]

    +......

    as explained


    µO approx. = 1/2E(O2) ( Put 2 oxygen atoms in a large supercell)
    Half of the total energy of O2 molecule gives the chemical potential of oxygen (µO).

  • David Foster added an answer:
    2
    How to find mobility of a particular system using VASP?
    Can I find the mobility from the vasp calculations?
    David Foster

    Note that the effective mass that is used in mobility calculations can be different than the effective mass associated with the density of states.  The former is given as m_c = 3/(1/m1 + 1/m2 + 1/m3) (see Sec 1.4.2 of Sze and Ng "Physics of Semiconductor Devices, 3rd ed.") instead of the usual geometric mean of m1,m2,m3.  Here m1 is the effective mass along principle axis 1.  The calculated effective mass using dense k-point lines and the second derivative is a low-temperature value.  It can vary from this (and becomes less well-defined) at room temperature when the CB/VB are not well approximated by a paraboloid all the way up to Delta E = kT.

    The effective mass is used in formulas for mobilities associated with phonon scattering and defect scattering, which assume that you know the elastic constant and defect concentration in the material respectively.  The elastic constant is easily looked up for known materials (or can be estimated using a VASP calculation).  Extrinsic defects densities come from doping defect densities, but intrinsic defect densities may be harder to look up (and they are a major project to calculate and the results can be off by more than an order of magnitude).  Calculate the mobility for phonons first, and if it is doped, calculated the mobility from extrinsic defects.  Then you can estimate the minimum intrinsic defect density at the intrinsic defect mobility becomes significant.  Hopefully, someone can tell you "the intrinsic defects in that material should be well below that room-temperature significance threshold".  I believe this is true in many cases. But for example, in CuInSe2, the intrinsic copper vacancies ~10^17 or 10^18 cm^{-3} (I believe) may be non-negligible at low temperatures, maybe even room temperature.  I have not looked at numbers enough to know at what temperature a 10^17 concentration would become important.

    In addition to the warning given by Buurma, there is the issue of the approximation of the functional you are using.  Band structure comparisons that I have seen between GGA and GGA+GW (more accurate, but very expensive to converge) look like the second derivatives agree well.  But I wouldn't expect a phonon-mediated mobility calculated this way to be more accurate than within a factor of, say, 5.

  • Kingsley Onyebuchi Obodo added an answer:
    6
    How can I calculate the guarantee of accuracy from the plane wave basis kinetic energy cutoff, Monkhorst-Pack mesh and Supercells atoms number?

    How to calculate the guarantee of accuracy from the plane wave basis kinetic energy cutoff, Monkhorst-Pack mesh and Supercells atoms number?

    I read

    "... using a plane wave basis kinetic energy cutoff of 800 eV and a 4 X 4 X 4 M-P mesh for k-points of the 40-atom supecells. These ensured an accuracy in the total energy of ~ 5 meV/formula unit. ..."

    How to get the "5 meV/formula unit"?

    Kingsley Onyebuchi Obodo

    In addition to previous statements on the k-point issue:

    To carry out a supercell calculation with same accuracy as your unit cell. The number of k-points should be scaled, otherwise one calculation will be more accurate than the other.

    Also, except you are introducing dopants or carrying out some modifications to your conventional unit cell, it is advisable to carry out your band structure calculation on your unit. You would have the benefits of faster calculation and more accurate band structure because you can increase the number of your k-points and other convergence criteria.

    Michael has provided a good answer on how to determine the kpoints for both unit cell and supercell calcations

  • Shuai Zhao added an answer:
    2
    Is the K-points path the same as the primary cell for a superlattice?

    In the first principle calculation, to plot the bands structure we need to calculate the bands along a K-points path through high symmetric K-points.

    Then, my question is if calculate the superlattice (2 x 2 x 2) bands structure, is its K-points path the same as that of the primary cell? If not, then how to set the K-points path for the superlattice?

    Thanks in advance.

    Shuai Zhao

    Dear Vaibhav, thank you very much for your answer.

  • Danny E. P. Vanpoucke added an answer:
    3
    Can you help me with VASP calculation of elastic moduli for hybrid functional (HSE,PBE0)?
    I'm trying to calculate elastic moduli for a system using a hybrid functional. My INCAR for the system is

    ENCUT = 600 # Energy cutoff for planewave truncation (higher -> more accurate energy)

    ISTART = 0 # 0 - Start new calculation
    ISPIN = 1 # Non-spin-polarized calculation
    IBRION = 6 # Calculation of elestic constant matrix (symmetry on)
    ISMEAR = 0 # Gaussian smearing
    SIGMA = 0.01 # Smearing width (eV)

    ISIF = 3 # Full relaxation / calculate full stress tensor

    NELMIN = 8 # Minimum number of electronic steps

    PREC = Accurate
    ADDGRID = .TRUE.
    LREAL = .FALSE. # Projection done in reciprocal space

    LWAVE = .FALSE. # Do not write WAVECAR
    LCHARG= .FALSE. # Do not write CHGCAR, CHG
    LVPOT = .FALSE. # Do not write LOCPOT
    LVHAR = .FALSE. # Do not write LOCPOT

    ## PBE0 (Hybrid functional)
    LHFCALC = .TRUE.
    PRECFOCK = Fast
    ALGO = Damped
    TIME = 0.4

    This works for regular LDA/PBE functionals, but not for hybrid functionals as VASP changes the k-point set on-the-fly, which causes an error:

    "internal ERROR in RE_READ_KPOINTS: the total number of non zero k-points in the full Brillouine zone has changed"

    Does anyone have any experience with this particular problem and, if so, could you suggest a fix or an alternative? Any help is very much appreciated. My version of VASP is 5.2.12

    (PS already tried IBRION=7 or 8, which is not implemented for HF calculations)
    Danny E. P. Vanpoucke

    If the issue is the k-points changeing, just switch symmetry off, in that case you will get a non-reduced sampling of the brillouin zone (which is more expensive, but should not die due to this error)--> ISYM = 0

Topic followers (558) See all