BANDSTRUCTURE CALCULATION using BURAI (GUI for Quantum Espresso) [TUTORIAL #5]

[@sofar524]

thank you for your hard work and explanation.. greatly appreciated!

[@ashiquzzamanshawon4944]

Hi, this probably requires some attention. The band structure of Si has an indirect band gap, between conduction band at gG and Valence band at X. That seems to be not the case for your calculation. Care to explain a little?

[@MuhsinTURAN-y6q]

I want to do bandstructure calculation for ReSe2 and ReS2,but the material project .cif files do not provide this data. What should I do?

[@PhysWhiz]

Check out some other publications that do calculations for the same material. Some of them may have provided the structures in the supporting information. Alternatively, you can even mail the authors (especially the corresponding authors).

[@raouzair8909]

if it's FCC why didnt we use Ibrav 2?

[@michaeljamescollins6800]

@Phys Whiz I am trying to learn QE and have some doubt about your result here. Si is an indirect semiconductor and in the literature the conduction minimum is at X point and the valance minimum is at Gamma point, this is even shown in your video. However, your DFT result shows both the conduction and valance minimums at the Gamma point which would describe Silicon as a direct semiconductor, which it is not. How can this be done accurateyly? THanks

[@PhysWhiz]

That is a great question. I agree, I didn't notice it at the time of making this video. The issue here is that of band folding. You see, I used 8 atoms of Silicon, .i.e. a conventional unit cell. More atoms means more bands. Usually band structures are plotted for primitive unit cells, i.e. 2 Silicon atoms, which will give you the expected result of X point CBM. This is a regular problem experienced in supercell calculations, and it is therefore usually advised to check for this phenomenon. Here is an article related to this: arxiv.org/pdf/1812.03925.pdf , another related link: www.physicsforums.com/threads/band-structures-with-non-primitive-cells.700772/. , www.tcm.phy.cam.ac.uk/castep/CASTEP_talks_06/clark2.pdf One can see that the conventional unit cell is 4 times that of primitive unit cell, therefore the Brillouin zone is 4 times smaller and hence the bands fold over through symmetry. Therefore, the band structure contains redundant information. Here is a good explanation from epubs.surrey.ac.uk/813175/1/PhD_Thesis_RossMaspero.pdf "There is another cost to using the supercell approach to calculate band structure and this comes in the form of band folding. Band structure for periodic systems is calculated in the Brillouin zone, which is the reciprocal representation of a primitive cell. However, as we increase the size of the real space structure to that of a supercell, the representative supercell Brillouin zone (SBZ) shrinks while the wavevectors we are interested in remain the same. Therefore, as the Brillouin zone gets smaller, the wavevectors start moving out the Brillouin zone but through symmetry are folded back in, losing its true k vector. How each wavevector is folded back into the SBZ is also not immediately straightforward and can be intuitively conveyed in multiple ways."

[@michaeljamescollins6800]

@@PhysWhiz thankyou so much for your thoughtful and fast reply!

[@abhijeetdas6279]

Respected sir Your tutorials are fantastic. I am a Ph.D. student from India and totally new to DFT. As a part of my research, I would like to incorporate a study on the resistivity behavior of Aluminium doped Titanium nitride for different compositions of Al in TiN. So, it would be very kind of you if you could guide me through the procedure. I have BURAI software with me. Thank you.

[@rajendraprasad6807]

What is the source of this gnu file you opened from your system

[@siphojonas5862]

like your videos..can you add calculations for TiO2 PLZ PLZ...

[@karreddularaja3756]

very nice lecture

[@shreekrishnatimilsina7529]

How can we calculate band gap from graph?? can you explain pleased

[@moulayyoussefraia5600]

How to calculate phonon , electric, structural, Optic, thermodinamic using burai Espresso ?

[@thanhhaile5810]

Dear Phys Whiz I have a simple question, Can you give me some information about this? The unit cell has a cubic phase. When I make supercell 3x3x1 (or 3x3x2) and 3x3x3. what ibrav value we can give for 3x3x1 and 3x3x3 supercell? By the way, how to increase the calculation time greater than a week for a large structure? Thank you very much.

[@vivekpendharkar5152]

Could you make a tutorial video for calculating stacking fault energy using Quantum Espresso?

[@valentinelun8958]

@phys whiz I think the crystal structure you were using is a simple cube so it got 8 si atoms, so I think we should change the input file in Burai to make it a fcc crystal system, then we should have 2 si atoms instead of 8. Then finally we can have the band structure comparable to the reference paper you brought in the video.

[@ManasSharma07]

I have mentioned this in replies to two or more comments already. There is an issue of band folding occuring here due to the 4 times volume of our conventional unit cell and 4 times the reference bands due to 4 times the electrons of a primitive unit cell.

[@valentinelun8958]

@@ManasSharma07 Right! Actually I selected fcc in the "cell" tag and there are still 8 Si atoms, do you know how to make the cell contains only 2 atoms? The reason I ask is I what to have a band structure looks exactly like the reference paper, which has 4 bands in the valence band. If you know how to make a band structure like that please let me know. Thanks!

[@PhysWhiz]

Make it ibrav=2 or whatever corresponds to fcc. Then in the atomic positions use crystal coordinates and 2 Si atoms: Si 0 0 0 Si 0.25 0.25 0.25

[@PhysWhiz]

You may find this video useful: kzbin.info/www/bejne/qnuunmmZe7Vjqac

[@valentinelun8958]

@@PhysWhiz Thank you !

[@rajendraprasad6807]

Where I can get gnu file

[@sarojthapa3873]

Hi, I am a new beginner in Quantum Espresso and trying to work on the bandgap and band structure calculations of CsPbBr3 perovskite materials and related doped compounds. Can you please suggest me a way to work on it and upload a video if you have any of those types of compounds?

[@learnin5minutes167]

Sir after doping interstitial atom like boron in metals do we do full structural relaxation or only ionic relaxation??

[@nilson5145]

I have a doubt, some band gap get with this program have not the accuracy, for exemple tha band gap of TiO2(anatase ) is 3,2eV in this program you got 2,1eV.

[@parameswariraju3119]

Sir, I am working on graphene supercell using burai, the band structure of supercell is not zero at K, how to make band structure similar to unit cell?

[@mounsifabbes5712]

what are the values inside espresso.band2.gnu file ??

[@johnconrad4439]

For the DOS calculation, where can i find the data file? I want to plot it in a different application instead of burai

[@science.ashutosh]

Hello.. Thank you for the interesting video. I am a beginner of BURAI. You have compared the x-axis of the graph with the different band point which corresponds to some number. What about y-axis, how to compare them and lebel the exact energy in eV?

[@PhysWhiz]

Hi, The energies are already in eV. I don't think anything else is needed. Although, you might want to change the reference point of energies as in literature it is common to take the Fermi energy as the zero level. In such a case just subtract this energy from all the energies to shift the energies down. This is trivial to do using Origin or even Excel.

[@muhammadwaqasahsan6374]

Band Calculation I have facing problem during band calculation. When I run the job SCF is converged and band calculations are done at about 551 points but after that BURAI donot show any response and Band energy graph did not appear. BURAI shows that job is running but even after 48 hours, no further calculations are done. Please guide me how to overcome this problem.

[@SAJALPHYS]

Sir, I Have installed Burai1.3 in my Ubuntu 20.04 PC, but I can't view the structures.

[@PhysWhiz]

Hi This is indeed a known problem with BURAI on Ubuntu which I haven't been able to solve. Only the Windows version works perfectly for me.

[@zakaryaehaman5888]

hi ... first of all thanks for all this videos ... My question is about adsorption ? how can i do Adsorption calculations using Burain ?? for example the distance between the adsorbate adatom and the graphene sheet... thank you

[@PhysWhiz]

Sorry but I don't have any experience with such calculations. BURAI probably won't be enough for such a calculation.

[@haaardcoooreee]

Silicon crystal has a diamond structure, as appears in the BURAI visualization, then... why in your input file ibrav = 1, that corresponds to a Simple Cubic structure, isn't? I have that question, because I work with Fe bcc and Fe fcc and despite the Burai visualization looks right, my input file has an Ibrav = 1, when I changed it to 3 or 2, my SCF calculation doesn't converge and quantum espresso launches this error: ERROR #41: S matrix not positive definite. so... Can I use the Burai input besides Ibrav value just making me sure that the visualization is the one I need?

[@PhysWhiz]

What I have shown here is the conventional unit cell of Silicon where the atomic coordinates ar given based on a simple cubic lattice, hence ibrav =1. This is why there were 8 atoms in my unit cell. For FCC structures one could use ibrav for fcc but then the atomic coordinates need only be provided for the fcc unit cell. For example if some structure has atoms only on corners and face centers then if you're using the ibrav corresponding to FCC then tou only need to provide the coordinate for 1 atom. And that would just be 0,0,0. This will help you run calculations faster as there would only be a single atom compared to 4 atoms for ibrav=1.

[@PhysWhiz]

Also you can refer to this video to understand in detail: kzbin.info/www/bejne/qnuunmmZe7Vjqac

[@haaardcoooreee]

Thank you so much!!!!!! I will watch the video, I hope it will help me to solve my doubt. I want to thank you for your channel, is very helpful!!!! Greetings from México.

[@johnconrad4439]

Hello sorry for bugging you like these I really have no one else to ask. There is something I'm really uneasy about. Everytime I do something with Burai that includes bandstructure calculations for a 3x3 monolayer MoS2 system (27 atoms), the entire project file ends up becoming 20 to 30 GB. For projects that only need scf and dos, the folder size is manageable even less than 1 GB but when I do bands it gets really really big. Is this normal or is there something wrong?

[@PhysWhiz]

I think it's normal.

[@johnconrad4439]

Oh okay thanks. I guess ill just have to figure out a way to manage my storage.

[@mel956]

out of curiosity, what does "Symmetry of Bands" do exactly?

[@PhysWhiz]

I guess you deleted one of your other comments, suggesting a mistake in my tutorial. And yes you're right i did make a boo boo there

[@mel956]

yea i noticed you answered the question at the end of the video haha

[@PhysWhiz]

If option lsym=.true, bands.x performs a symmetry analysis. An additional file sibands.dat.rep is generated, containing information on symmetry labels of the various bands.

[@anishdas39]

Hi, i am a beginner in espresso, your video is very help full, but i have some question or doubt. 1) (Minute: 15.47) Your eight atoms band calculation very much replicate the two atoms bands of the literature but what about the Fermi energy? In two atoms bands calculation of the literature the Ef is defined well (at the top of the highest occupied valance band) and is scaled to zero but in your case, the Ef isn't quite defined. If i consider though the ZERO eng. level as Ef, it is deep inside the valance band, how does it become so? Though it is possible for degenerate P type material, so is it p type? I am a basically a self learner so may be sometimes my question looks nondescript, please pardon me. 2) Request: Can you make a tutorial video of calculating EV curve? Because generating EV curve is my point of interest. If i am able to generate EV curve then further i can use that potential/force field for MD simulation to calculate thermoelectric parameters.

[@jesuscanasfernandez8494]

I am not familiar with qe but generally in this type of calculations the absolute energy is arbitrary. In case of the simulations you have seen in literature they have just fixed their parameters in order to set top of the valence band to 0.

[@johnconrad4439]

are BURAI calculations LDA or GGA? sorry you may have mentioned this somewhere already but I cant find this.

[@PhysWhiz]

Actually burai just uses quantum espresso, and QE basically decides LDA/GGA based on the Pseudopotentials. Purdew zunger PPs are LDA i believe. PBE are GGA PPs. For hybrid calculation though you should use like PBe PP and specify explicitly in input_dft variable. For more details look for input_dft variable in system namelist in pw.x input file description.

[@johnconrad4439]

@@PhysWhiz thanks so much for answering. if i may ask a bit more, suppose i do dope substitution for example on CdTe and i replace one Cd atom with Mn. do i still need to do a relaxation or optimization or can i directly proceed to doing the scf calculation? appreciate greatly your answers

[@PhysWhiz]

You definitely have to optimize the geometry

[@abishekkhatri7368]

@Phys Whiz i get this error when i run burai calculation Note: The folling floating-point exceptions are signalling:IEEE_DENORMAL cal you please help me understand what is error means

[@PhysWhiz]

That is not an error. The calculation must have been successful.

[@PhysWhiz]

github.com/marvel-nccr/quantum-mobile/issues/52

[@abishekkhatri7368]

@@PhysWhiz thankyou sir

[@Jatin19902]

Hi, I am studying a graphene nano-ribben with ripple in it. The system contains 120 atoms. In scf output file it shows 240 eigenvalues for any given k point. Can you please help me to decide what should be the nbnd value in bands calculation with pw.x following the scf calculation. Thanks.

[@PhysWhiz]

Carbon has 6 electrons in total. In codes that use Pseudopotentials, only the valence electrons are used in DFT calculations. You probably had a pseudopotetnial which considers the 2 1s electrons as the core. The remaining 4 electrons are considered valence and used for DFT calculations. Therefore, for 120 atoms there would be 480 electrons. Since there can be 2 electrons in 1 energy level(orbital), therefore we have 240 occupied orbitals (energy levels). You see the energy values corresponding to these in the scf output file. The nbnds basically tells quantum espresso how many bands/orbitals you want the energies for. For example an scf calculation with nbnd >=241 in your case will give the energies of conduction bands. This will enable you to find out the band gap from the scf output file itself. For a bands calculation, the nbnds tells quantum espresso, how many band syou want to see in your band structure. So if you use nbnd = 20, then there would be 20 conduction bands in your bandstructure. Hope it helps you with your doubts.

[@Jatin19902]

@@PhysWhiz Thanks for your informative and quick replies. Can you please quickly help me with my following doubts. When you say "For example an scf calculation with nbnd >=241 in your case will give the energies of conduction bands."; you means to say this will give me the energies of valence bands(240) and a conduction bands(1), right? And my second doubt is when you say "So if you use nbnd = 20, then there would be 20 conduction bands in your bandstructure"; Won't be those 20 bands will be valence bands in my case? Moreover, for my given case as I have 240 occupied orbitals then what should be my nbnd in bands calculation. Shall it be anything higher than 240. If yes then won't be it too many bands on one plot which will make it harder to understand what is happening on the bandstructure plot. Can I change the nbnd value to make the plot less messy and also see the conduction bands and valence bands at the same time. Thank you!

[@ManasSharma07]

@@Jatin19902 "you means to say this will give me the energies of valence bands(240) and a conduction bands(1), right? " --Yes exactly. "And my second doubt is when you say "So if you use nbnd = 20, then there would be 20 conduction bands in your bandstructure"" --Yes, my bad. I meant that there will be total 20 bands, not conduction bands. "If yes then won't be it too many bands on one plot which will make it harder to understand what is happening on the bandstructure plot." --No, even if there are a lot more bands than you need, you can always zoom in to see what's happening. Of course, it depends on whatever bands you are interested in. For example for a semiconductor, I would usually be interested in both the valence and conduction band. Therefore, I would give an nbnd value such that there are a few conduction bands drawn as well. This will, let me show the band gap in the band structure. It would also be useful to how doping or some other parameter affects the broadening or narrowing of bands. Since, all the valence bands are calculated, in the band structure file, and I might not need all of those. I may crop the band structure to focus on the outer most valence bands. But again, it all depends on what I want to see. In your case, since 240 bands are already going to be making the plot messy, I don't think having 5-10 conduction bands would make it messier. Unless, you're not at all interested in the conduction bands. I don't think there is a way to prevent the calculation of the valence bands, but what I generally do is, I just zoom in to the band structure cropping the lower valence bands which are not relevant to me.

[@Jatin19902]

​@@ManasSharma07 Thanks a lot again for clarifying my doubts! Can you please help me to understand something that I am struggling at. When I don't specify nbnd value in SCF calculations. Then for the same system if I run the calculation by changing the convergence threshold(conv_thr under &ELECTRONS block) then it gives me different value of "number of Kohn-Sham states" in SCF output i.e. 240 lets say in case A and 288 in case B for same system consisting 120 Carbon atoms having 480 electron. The pseudopotential used is having 4 electrons/atom used in DFT. And I assume number of bands printed for each K point in SCF output is equal to "number of Kohn-Sham states" as I can observe from all the SCF output files I have. I am not able to understand why is the number is not same i.e. 240 for both case A & B. Why I got extra 48 in case B. Is it due to degeneracy of the orbitals? Considering that, would that change the number of HOMO and LUMO in case B from as that of case A. Would HOMO will be 240th orbital and LUMO will be 241st orbital for both case A and B. OR it would be that for only case A but for case B HOMO will be 288th orbital and LUMO will be 289th orbital? Once again thanks for the resources!

[@PhysWhiz]

@@Jatin19902 That is indeed weird. Did you increase or decrease the threshold? I would suppose if the results seem weird then try with a tighter (smaller) convergence criteria. Also, it could be due the fluctuation of occupations near the fermi level. Maybe the conduction bands are really close to the fermi level like a metal. In such cases, the occupations may fluctuate leading to sci convergence issues as well. Therefore for metals, smearing is advised to have smooth occupations for density integrals. I definitely haven't faced such an issue before so am not sure about what could be causing it.

[@kemalaziz9696]

Was the pun intended at 4:00?