DOS - Science topic

In PDOS calculations, TDOS may appear less than the sum of individual atomic contributions due to the following reasons:

1. Projection basis limitations: PDOS is computed by projecting wavefunctions onto localized atomic orbitals, which are not complete in the full plane-wave basis. 2.Overlap issues: Delocalized states (especially conduction band states or from hybridized orbitals) may not be fully captured by atomic projections. 3. Numerical shortness and cutoff errors: Low-projection weight states may be filtered out below a threshold.

Therefore, TDOS is typically greater than or equal to the sum of PDOS, and discrepancies are expected. This does not imply an error but is a known limitation of projection-based analysis.

Dear Sushil Kumar I have resolved it by scrupulous fitting of following blocks:

PAO.Basis, PAO.BasisSize, PAO.SplitNorm, and LDAU.Proj (for Co/Fe atoms treatment). Finally, it was shown to provide expected results, same as in the VASP plane-wave package. So, my advice is to play with parameters carefully.

In the literature result you quote, the energy zero was arbitrarily set at the top of the valence band. It is not the Fermi level, since there is still one f

electron in the lowest impurity band. What counts here are the relative energies of the components of the DOS, and your values concord with the ones from the literature.

Thank you very much Jürgen Weippert

Hello Tercio Paulo Xisto ,

You need to remove few lines in the generated .dos file. If you need more info please message me. i am unable to message you.

Hi Megha Sharma

I am having the same issue as you with a very "spiky" DOS. Did increasing the k-points work for you? I am using twice the k-point density as when I did my relaxation calculations but my DOS still isn't as smooth as I had hoped.

Regarding the out-of-memory error, if you are working on a computer cluster, the issue can often be solved by either increasing the number of nodes and/or reducing the number of cores per node (so that some cores are used for memory only). This link has more information on the topic:

https://www.nsc.liu.se/support/memory-management/

Best,

Anders

First of all, we have to consider the possibility of averaging a stack of several captures (>10) to eliminate temporal noise. So when I say "image", I mean the average of the stack. In this way, we can minimize the problem of spatial non-uniformity. What we have to do is:

1) to obtain an image in darkness, I'll call it DSNU;

2) to obtain an image in front of a Flat Field at the end of the scale in the dynamic range of the instrument, but making sure that there is no pixel in the image that enters into saturation; I'll call this FF;

3) to obtain PRNU = FF - DSNU;

4) to obtain the average value of PRNU, I'll call it PRNUm;

5) and considering that we already have the reference image, a raw image, which I'll call RAW;

6) finally you can obtain your corrected flat field image (Icorr), assuming a linear behavior throughout the entire dynamic range of the sensor, as follows: Icorr = (RAW-DSNU) x PRNUm / (FF - DSNU).

Regards!

Well, the different lattice structures of the transition metals canbe explained from the DOS moments theorem (see Pettifor 1987 or the 1993 book by Sutton). That, however, does not explain intermetallics.

1) You haven't attached a graph.

2) In simple cases, the UPS with subtracted secondary electrons directly depicts the DOS. In the supplement to this:

you find a direct comparison of UPS and a primitive theoretical DOS which was generated by putting Lorentzian functions over DFT orbital energies. As you can see, the agreement is already quite decent on this simple level. Deviations may be due to the absolute orbital energies, cross section effects and eventually the baseline subtraction also plays a role.

3) If you go towards materials with a higher crystallinity, a single UPS is no longer sufficient for that because angle resolution plays a massive role. In that case, one spectrum only depicts the dispersion relation over one point of the Brillouin zone, so you need a full ARUPS.

El Mehdi Bouchtaoui

SeinFit appears to be an older fitness software program, and it's possible that a DOS version is no longer readily available for download.

Hi Sannyal,

i think i have simillar problem have you found out the reason?

These terms come into play when the system being studied has unpaired electrons, leading to a need to distinguish between the two spin states. This is typical in systems with open-shell configurations, such as radicals, transition metals, or molecules with odd numbers of electrons.

Change in number of electrons can occur if u have changed the pseudopotentials used.

To plot the Density of States (DOS) up to 20 eV, you need to modify your input files to ensure the energy range covers up to 20 eV. First, open the .in1 file and set the energy range line to include values up to 20.0 eV, then save the file. Next, open the .int file and make sure the energy range is also set to 20.0 eV, then save this file as well. Run your DOS calculation using the appropriate commands for your software (e.g., x lapw2 -qtl -up and x lapw2 -qtl -dn for WIEN2k). After obtaining the DOS data, use plotting software such as gnuplot or Python's matplotlib to visualize the DOS up to 20 eV. This ensures your plotted DOS includes the desired energy range.

Bonjour Farook,

J'ai transféré votre message à ma doctorante KHOULOUD, c'est elle qui a fait le calcul des propriétés vibrationnelles du germanène avec le VASP

Hello Protsan Dhungana

You can save the DOS data from GaussSum software, which you can plot in the Origin software. Please check out the website and documentation of the GaussSum software.

To run VASP in non-collinear spin mode (which is the case using spin-orbit corrections), you need to use the vasp_ncl executable as the wavefunctions will be spinors in that case so that you can't use the standard executable.

Hi,

To give you a rough idea on how to proceed:

1) build a clean surface (graphene in your case) and run a calculation with it;

2) add the Li atom(s) and repeat.

Notice that unless you what a full coverage of Li atom of the C surface, you need to have a supercell made of graphene unit cells to reduce the ratio Li/C.

If you are proficient with Python and Jupyter, I recommend ASE (atomistic simulation environment) as a tool to generate the both the pristine graphene, the supercell and then add the Li atoms. ASE will provide the atomic position and the lattice parameters that you will need to include into the Quantum ESPRESSO input. (Indeed, you can create the input directly within ASE.)

To complete your calculation you will need to:

1) Run a SCF calculation to determine the electronic ground state density. This step requires also the convergence of the simulation parameters (energy cut-offs, first Brillouin Zone sampling).

2) Run a non-SCF calculation for the band structure on a path

3) Run another non-SCF calculation for the DOS and PDOS on a mesh of the first Brillouin Zone.

You have to repeat the above steps for each of the configuration you want to investigate (i.e. changing the Li atom positions and their number).

I hope this helps,

Roberto

Yes, you should. Because step 3 is a non scf calculation, which must start from a converged wavefunction and charge density, otherwise the non scf calculation of DOS will start from a random wavefunction, which is totally different from wavefunction from step 3. Therefore, you should use wavefunction and charge density from step 2 to ensure that your DOS calculation is exactly the same strcuture and the same wavefuction, the same charge density to step 2.

Hi Roberto D'Agosta

I tried reducing the k-point to 5 and 4 and I am still facing the same problem. I'm using virtual machine, so I think this is the cause of the problem. I will try to find another method to calculate DOS. I really appreciate your help, thank you so much.

Mohamed Ali Lahmer Thankyou Sir for your reply. Along X-W how can we take the K- points from the data file because in the X-W direction if we consider the heavy hole band (band 1), then we get same value of effective mass if we take the same number of K -points for non-linear fitting.

I encountered the same error computing the PDOS. I bypassed the error by setting the variable 'nosym' equal to '.true.' in the &SYSTEM card in both the SCF and NSCF calculations. With this, I was able to get the PDOS data and plot.

This error is attributed to broken symmetry in hexagonal systems, as Paolo Giannozzi said in the pw_forum on narkive.com.

For most calculations, you should only keep the CHGCAR since DFT as a theory relies on Electron Density. In fact, to conserve storage usage on the HPC system I use, I switch off printing the WAVECAR file by using the tag LWAVE = .FALSE. If you need to restart a calculation, simply use the CHGCAR file.

For some calculations, however, it is necessary to keep your WAVECAR files. Some exceptions I can think of at the top of my mind calculations are GW and optical properties calculations. Here are some examples from VASP's manual.

Dear friend Abdullah Al Imran Ifrit

The specific defect parameters you are requesting, such as energy band gap, electron DOS in the valence and conduction bands, electron affinity, intrinsic doping, NTA, NTD, WTA, WTD, WGD, EGD, mobility of electron and hole, and permittivity, can vary depending on the specific conditions and experimental setup.

It's important to note that providing accurate and up-to-date defect parameters requires access to detailed research data and characterization techniques specific to PEDOT: PSS. Therefore, it would be advisable to refer to published research papers, and/or materials databases in the field who provide expert information in the characterization and simulation of PEDOT: PSS. They will be better equipped to provide you with the most accurate and relevant defect parameters for your simulation study.

I am curious to know and explore more with you.

Dear Trivanni Yadav,

You should specify that your questions are about VASP so people can help you more directly. I can indeed only answer your first question, but I believe the other two should be in VASP manual…

In general, we do not need to specifically start from a SCF charge density. You can start from whatever proper density (that is, a definite positive function integrable to the number of electrons in your system). The way you obtain that charge density somewhat defines the level of accuracy of your calculation. For example, if you had access to Monte Carlo-calculated densities, those would provide an excellent starting point for further calculations. You could construct a density starting from a linear combination of the occupied atomic orbitals. Clearly this density would be a poor starting point as we know it neglects important correlation effects.

In the case of DFT, to obtain the charge density you have to solve the non-linear Kohn-Sham equations for your system. An efficient, robust, and general method to solve non-linear equations relies on self-consistent loops. The same method, for instance, can be used to solve the Hartree-Fock equations. So there is nothing inherently magic into the “SCF-generated charged density” if not a general method of numerical solution.

I hope this helps clarifying your doubts,

Roberto

@Xu Xuanwen did you find out how to calculate Fermi energy and bandgap from the graph you provided above?

It is semimetalic. The gap is on the Fermi level. The Fermi level by the definion is always located at the top of valence band (in VASP), not in the gap.

The relationship between quantum capacitance and DOS in the non-degenerate limit depends on the electronic structure of the semiconductor and can be approximated using the Thomas-Fermi approximation or more sophisticated models. To calculate the DOS from the quantum capacitance, one needs to solve for D(EF) using the appropriate model and input parameters.

One way to approach the problem is to use the Thomas-Fermi approximation, which assumes that the electronic structure of the semiconductor can be approximated by a local density of states that depends only on the electron density. In this approximation, the quantum capacitance is given by:

Cq = πħ^2D(EF)/2me*(dEF/dn)^2

where D(EF) is the density of states at the Fermi level, me* is the effective mass of the electrons, and dEF/dn is the derivative of the Fermi energy with respect to the electron density.

It is relatively simple. If you integrate the DOS (from -infinity to the the Fermi energy) in both, the majority and minority spin sectors , their difference gives the global magnetic moment per unit cell.

If you have the DOS resolved in species (local DOS) then you can obtain the magnetic moment per type of atoms (or orbitals).

However, you will not get any information regarding the couplings between the localized moments from the DOS.

The gap you can just get directly from the dta, and you can calculate effective mass, see https://github.com/lucydot/effmass, from which you can compute Nc/Nv.

The other quantities require additional calculations, e.g. a typical approach with mobilities is to use deformation potential theory, and an additional self consistent calculation for the electron affinity, though electron affinities are typically quite bad with DFT unless you're very careful with how you compute them.

In the most general terms, absolute energy is meaningless; only energy differences are observables.

Choosing the fermi level as our "0" is very convenient when studying a single material because it is one of the things we can easily calculate.

Vacuum energy is also a valid choice of zero, and as you say is more appropriate for comapring band alignment between different materials.

The reason vacuum energy is not calculated "as standard" is often due to the way we have designed most DFT codes - modelling a bulk solid in periodic boundary conditions has *no vacuum* and therefore an additional slab calculation (i.e. a material-vacuum interface) needs to be calculated to find E_Fermi relative to the vacuum energy.

In the plane-wave codes that many people use, vacuum costs as much as matter in terms of minimisation, making these calculations *not cheap*.

In short; yes, you are right, but if you aren't comparing different materials in your project, vacuum levels are a lot of extra work for no benefit!

Hope that helps!

You are most welcome, Dear Smrutiranjan Mekap

Kind Regards.

Vipul Kumar Ambasta By increasing q point can solve (fast forwarding solution). Let's say nq1 = 5/6, nq2 = 5/6, nq3 = 5/6

Further

1- I have doubt in smearing factor what you used (Gaussian).

2- I don't think so your digits of KPOINTS considerable. It should be more dense.

3- I never show ecutwfc = 35, please re verify.

4- Please once choose separate cell parameter and ibrav = 0 for doing vc relax.

Well, if you have a structure file, you would have to do DFT or another quantum chemical method next. But there are plenty of papers out there which have already done that, so you could just ask the authors whether they could provide energy output files to you.

It indicates the presence of unpaired electrons in the system and it results in the net magnetic moment of the system.

Dear Sharad:

You can benefit from this valuable handbook about your topic, which I attached below ...

I hope it will be helpful..

Best wishes..

Hi, Sumit Kukreti

I think you have two questions, (1) How to calculate the DOS contribution of different sub orbitals of d in WIEN2k, (2) How to find the signature of Jahn-Teller distortion in the density of states.

So, answer to the 1 part, (1) In WIEN2k, you can use different ISPLIT values in the case.struct file to calculate contribution from different orbitals, ISPLIT = 8 will provide all the different contributions.

Now, comes to the second part

Even, only the contribution of d-eg, d-t2g orbitals can tell you about the JT distortion. In order to check that, you must look at the DOS of d-eg and d-t2g orbitals. If they have contribution corresponding to the different energies, then there is an effect of crystal field splitting. Now, if you noticed the instead of a single peak of D-eg or D-t2g orbitals, if they are more than one peak. It will provide you the information that degeneracy of these levels have been lifted. This phenomenon is known as the JT Distortion.

You can look into the below references and references therein for more insight.

Thanks.

By now, I found the problem:

If you specify a value for NSW>1, the DOS is divided by this number.

Actually, I just realized that I discovered this problem already a year ago and clean forgot about it:

This will help you.

(equation -3) David Ansi

Hello Muhammad,

You need to assign each atom spin in the structure. Then you you need to perform scf and nscf with spin true option.

Regards,

Enamul

It is very difficult to give suggestions with little information. Please improve your question.

Dear Sanjeev,

Did you clarify your query regarding quantum capacitance. I also want to calculate QC from DFT for one of my system according equation (3) of appended article.

I emailed author of this article, he replied me properly each step but still I could not calculate QC from my DFT results due to one another doubt. My doubt is that how does take DOS(E) and E values from DOS plot for solving equation (3) of above given article. If you will share your email_id, I will forward that email to you.

Thank You

Maybe you can refer to https://mattermodeling.stackexchange.com/questions/3856/how-to-determine-la-ta-za-zo-and-lo-to-by-phonopy.

see

As per the definition of extrinsic semiconductor there is formation of two type of semiconductor based on dopant (1) n-type, when the dopant is electron rich in which case the fermi is shifted towards conduction band, while (2) p-type, when electron deficient dopant shift the fermi energy towards the valence band. So aluminium doping leads to formation of exrinsic semiconductor.

Maybe your material is Ferremagnetic

I would like to convey my sincere gratitude to you Sruthil Lal S B for your comprehensive explanation of possible reasons.

I am working in a cluster where scaLAPACK is already installed. My job not terminating at the end in fact it stops at the beginning itself. It requires memory of around 1Gb per core ( I percept that, as the job terminating from the beginning it is showing less memory, apologies if this statement is wrong!!).

I have tried to increase the NPAR value up to NPAR= 32. this time the job runs up to 4-5 iterations. Then again the job was terminated with the same error.Please help more.

Thanks & Regards

Mousam

The input file you send me seems to be using smearing occupation and not fixed, with which the calculation runs fine. But the major mistake I found in the input was the k-point number is very low. use 6x6x1 k-point and the calculation will run fine with fixed occupation. Nevertheless, you should be using smearing occupations in the case of graphene as it is a zero-gap semiconductor.

see this:

i also try typing "help" in the command prompt to get all the commands . i find it very useful command.

Dear Mirza,

The PDOS is a projected density of states. In quantum simulation, with VASP or CASTEP, etc., PDOS means the site-projected DOS, where the information about the different contributions of different orbitals is computed. The difference between DOS, local DOS (LDOS) and PDOS was discussed in the following link:

If your film is thin enough to still have some electrical contact with the substrate, you can probe the VB-DOS by UPS and the CB-DOS by IPE (inverse photoemission, the machines for that are pretty rare, though).

Alternatively you can try STS (scanninc tunneling spectroscopy), but that requires even more conductivity to be left in your film than UPS and IPE.

You can use kresolveddos tag in Quantum Esspresso, all of required information you can find in the following link: https://www.quantum-espresso.org/Doc/INPUT_PROJWFC.html

Is your system metallic or non-metallic? What smearing are you using for the Kohn-Sham states? Have you computed the filling of the d- and f-states, or are you assuming they are filled based on their location in the periodic table?

In VASP you need to set LSORBIT tag to .TRUE. to take spin-orbit coupling into account in your calculations. Enabling this tag calculates the energy of the system for various direction of the magnetic moment.

Further, if you are particular about any spin quantization axes, you can set them to your choice with SAXIS-tag. Hiba Aljayyousi

Dear Parminder Kaur

Welcome,

The density of state DOS energy distribution in amorphous and microcrystalline materials are given in the thesis at the link:https://uwspace.uwaterloo.ca/bitstream/handle/10012/3884/Kyung-Wook_Shin_Thesis_format_revised.pdf?sequence=1&isAllowed=y

It contains also input files for thin film FET transistors.

Normally the DOS parameters are obtained by curve fitting of measured characteristics:https://uwspace.uwaterloo.ca/bitstream/handle/10012/3884/Kyung-Wook_Shin_Thesis_format_revised.pdf?sequence=1&isAllowed=y

Best wishes

You can try SREM, the simplest, robust, and efficient atmospheric correction method. Let me know if you need any help.

FYI:

You will find what you need on the website of Ronald Bakker:

You can download there the software package CLATHRATES for Mac and Windows.

Good luck,

Matthieu

If you just want to estimate the net electronic spin polarization, all you need to do is to take the values of DOS_UP and DOS_DOWN at the Fermi level and then put them in the following expression:

SP=ABS(DOS_UP-DOS_DOWN)/(DOS_UP+DOS_DOWN)

Above ABS(DOS_UP-DOS_DOWN), means the absolute value of the difference between DOS_UP and DOS_DOWN.

That's funny Eric, I sent you a similar shell script I wrote using awk like 4 or 5 years ago. Did you figure it out yet? It's on my github.

Thank you very much sir.

&PROJWFC

prefix="Ag2O"

outdir='./TMP_DIR2/',

ngauss=0

degauss=0.0002

Emin=-20.0

Emax=20.0

DeltaE=0.01

filpdos="1-pdos"

/

After SCF calculation there is a package call projwfc.x for PDOS and dos.x for DOS. This input can be followed. The outdir should be same as the one in SCF calculation.

Dear Yousif Shoaib Mohammed. I have gone through the attached files you recommend but still I need your kind support. I am working on Multiferroic materials (BiFeO3) using Quantum Espresso but i failed to calculate the polarization of the material, thanks.

Hi would you mind send me exact link i need it asap

Hello

I hope you have created the unitcells with different volumes (0.95 ... 1.00 ... 1.05) . Inside each unitcell, you have to create the supercells (ex- phonopy -d --dim="2 2 2"). This will create many POSCARs base on the symmetry of your system along with a SPOSCAR file. Now, you have to use this SPOSCAR as POSCAR to run a simple DFT calculation. The OUTCAR will have the energy and volume value. Repeat the same procedure for all the volumes to get the e-v.dat file.

If you need a dense k-mesh to calculate thermoelectric properties, you can use the utility program BoltzTrap which is compatible with most of the DFT codes such as WIEN2K, VASP, QE, etc.:

Alternatively, you can try Wannier90 which does this for you using maximally localized Wannier functions in real space and without any k-mesh sampling: http://www.wannier.org

I personally prefer the latter.

Dear Monalisa Nur Ramadhan

The first and most important parameter to set in your input file is the value of Fermi Level. If you want to consider n-doping set this value to the energy corresponding to the bottom of the conduction band. On the other hand if you want to study the effect of p-doping fix it at the top of valence band. The parameters, deltae and ecut, then allow you to vary the chemical potential around the Fermi level at regular steps (delte) within your desired range (ecut). Alternatively, if you use the latest version of Boltztrap you can directly specify the doping concentrations in the input as well (still you need to specify Fermi level). For example to compare the Seebeck resulting from doping 1x10^20 cm^-3 electrons with that of hope doping of 2x10^20 cm^-3 you can add the following commands to the input:

2

1E20 -2E20

Above, The first line is the number of fixed dopings and the second line defines the corresponding values. For more information, please refer to the manual of Boltztrap (attached here)

I had the same problem with starting the analysis.

Hi Dr M.A. Hadi,

I think it is better if I add something. My preceeding answer was a bit uncomplete.

If you need to know the strength of interatomic bonds (Eb) of a crystal from the DOS then you need first of all to derive the fundamental Eg energy bandgap (smallest one) because it is this energy which reflects more or less the bonding strength Eg=Eb.

To determine the energy bandgap from the DOS calculted e.g. from ab initio computations then you need to know where are your antibounding conduction band states and where are your bounding valence band states.

Sometimes it is given directly by the software including the value of the energy gap.

If not then you need to compute the energy band structure and then it is easy to determine electron-like states with positive E(k) energy dispersion (positive second derivative of the E(k) energy dispersion as a function of wave vector k or a parabola directed upwards in energies) from hole-like states with the reverse situation means negative E(k) energy dispersion (negative second derivative of the E(k) energy dispersion as a function of wave vector k or a parabola directed downwards in energies).

Then all the E(k) energy curves showing electron-like behavior constitute the conduction band, and the E(k) curves showing hole-like behavior constitute the valence band.

Then in the Brillouin zone, it is easy to find the highest valence states (means the highest maximum) and the lowest conduction states (the lowest minimum) and the energy difference between these two points is the fundamental energy bandgap Eg you are looking for.

To be sure, you must plot then the E(k) energy band structure and the DOS on the same diagram with E(k) on a left diagram and E(N) where N is the density of states on a right diagram. Both diagrams have same ordinate as the energy and different abscissas: the abscissa is k (Brillouion zone wave vector) in the case of E(k) variations and it is the DOS N in the case of E(N) variations.

Once done, it is easy to see that energy bandgap in the energy band structure (left diagram) is almost exactly the same as for the DOS on right diagram. This is the proof that you derive correctly the fundamental energy bandgap Eg.

Now as to the shape of DOS as a function of energy.

There are usually several peaks in the conduction band and several other peaks in the valence band with weighted contribution of s, p, d and f states (if any).

- The |s> states are of spherical symmetry and are delocalized states in real space. Correponding wavefunctions are strongly localized in reciprocal k space.

The DOS peak corresponding to such states will be rather broadened (large width and small peak) because of strong ovelap of neighboring s-states.

- Conversly the |d> or |f> states are strongly localized in real space i.e. strongly delocalized in k-space, and corresponding DOS peaks will be rather narrow (very thin and long) because of very small overlap of neighboring such d- or f-states.

- the |p> states are intermediate between s- and d- states in terms of localization but closer to s- than to d-states.

In the simplest materials, usually:

- The |s> states give the dominant contribution to antibonding conduction states and the lowest conduction band DOS peak is rather a smeared peak (broadned peak).

-The |p> states give the dominant contribution to bonding valence states and the highest valence band DOS peak is a bit less broadened than the conduction band one.

However, this simple picture which is the fundamental one may change with some new features as strong hybridization of sp states with each other and with |d> and |f> states on the other hand.

This hold in covalent materials with increasing energy bandgaps with mismatched atoms (a mixture of heavy and light atoms) .

This is even more the case in insulating materials with large energy bandgaps and ionic bondings.

I hope this could help you.

It is long but I could not do it shorter.

Best regards,

Professor A. Kadri

Given that you want to measure performance and take packet traces, I would implement the topology and software shown in Figure 1 in the paper of Arianit Maraj mentioned above. To grab packet traces, I would put a system on the switch between the pen test machine, port mirror the victim machine's port, and then something like tcpdump or wireshark to capture the packet traces. The paper describes web page performance measurement software you can use.

hybrid functionals usually only allow scf calculations. n-scf, dos , bands etc calculations are not yet feasible as far as i know.