Dear Zhiwei, its tesseral spherical harmonic, so it is consistent with all other usage in the code. best wishes Michael PS: Remember that if you use the switch lmirep the columns are done with the adapted projections!
Okay Zhao, I played around with your input cards and came to the following conclusions: There exist two magnetic states which elk can converge in the case of no-ls coupling. These are a high spin state (4+ mub on Fe) and another low spin state (1 mub on Fe). This depends a little bit on the U you apply. Increasing your U always induced the second state. I think now once you switch on LS coupling you may modify the bands structure and hence the low spin state becomes favored, which moment even becomes...
Dear Zhao, could you try linear mixing instead so set mixtype 1 and then reduce the mixing betamax 0.3 best michael
Dear Zhao, I would first check if you need a dft+u calculation here. Is this system an insulator? How large is the DFT-noSC gap. If it is an insulator and the gap is already small, it might be closed by the LS-coupling and then it will be hard to find the insulating AFM solution. If this is not the case, I would suggest checking to start with larger fields instead of 0.01 maybe start with 1 and set reducebf 0.9 so the field is kept longer on. This might help to achieve convergence of the AFM state....
Dear Sabrina, first thanks for your INPUT. Here my accessment the unit of the Chi2 should be distance/voltage so in CGS we have this in statV/cm and the conversion goes by 5.8300410-8 cm/statV== 1 (BohrRadiusElectroncharge)/HartreeEnergy so then to go from cm/statV == 0.0000333564 m/V which is different than 4 [Pi]/3000=0.00418879 so how does this value come about? In any case also followed your suggestion with the exciting tutorial and the slides from CECAM and indeed it looks like that the units...
Dear users, I guess there was somehow this question already but somehow I could not resolve it for myself. So what is the unit of the nonlinear coefficients in elk given in CHI_2WWW_abc.OUT? So, I thought there are given in pm/V just in atomic units which would mean that we would have a0*e/Eh (a0= Bohrradius e=electroncharge and Eh Hartree energy) as a conversion factor, which would mean to go to SI we have to multiply by 1.9446. However, this seems to be way to small as if I take the GaAs example...
At least the 0 displacement offset sounds for me like a symmetry issues. Could you shift off your k-grid to the magic center point via vkloff 0.25 0.5 0.625 and then see if the curve becomes smooth best Michael
May I ask how your AFM setting looks like ? best Michael
May I ask how your AFM setting looks like ? best Michael
Dear Wonjae, I think this is not yet a code feature, as he either constrains the moment (all three components) or nothing. So I am not sure what you want to achieve here but wouldn't be better to just initialise the moments with the specific directions and let the rest relax completely for itself ? So remove mommtfix. best Michael
You can just leave them out and it should be fine just check in info out.
You can just leave them out and it should be fine just check in info out.
Dear Wonjae, you specify the first four moments as mommtfix 1 1 0.2304673996 4000 4000 1 2 -0.2304673996 4000 4000 1 3 -0.2304673996 4000 4000 1 4 0.2304673996 4000 4000 which is also above 2000 as a norm an hence the code will set the constrained field to zero! I assume you want to ste the 4000 to zero thus mommtfix 1 1 0.2304673996 0 0 1 2 -0.2304673996 0 0 1 3 -0.2304673996 0 0 1 4 0.2304673996 0 0 should work. Also, the moments you do not want to specify can just be not set and the code in the...
Dear Zhiwei, did you try to change mixing to mixtype 1 ? Although, definitely slower this may at least give convergence. best Michael
Dear Harry, the reason being for this suggestion was that often (in my experience) such strange elk results are resolved by moving to k-grid to a non-symmetric center point. Recently, we had here the case that the equation of state curve did came out smooth and only by moving the mesh changed. Let me try your calculation and see if I can do something. Although the spin-spiral extension also has been a mystery so far for me. best Michael
Dear Harry, could you check if it is possible for you to shift the k-mesh via vkloff to 0.5,0.625,0.05 so its off symmetry in some cases I had sucess to get rid off the strange curve behaviour. best wishes Michael
Dear Harry, could you check if it is possible for you to shift the k-mesh via vkloff to 0.5,0.625,0.05 so its off symmetry in some cases I had sucess to get rid off the strange curve behaviour. best wishes Michael
Hi, there is a nice toolbox for doing this see https://pypi.org/project/cif2cell/ which also supports elk best wishes Michael
I think its easier to convert the final result, rather than to rewrite the code
No problem are you using GGA currently? Could you check if this gives you an insulating state? If yes, then the you always will have a net moment of about 4 muB per site, depending on the MT. You also may consider switching off spin orbit for testing as this will not change to much here. best Michael
Dear Hanabi, Firstly, GaFeO3 is considered as a correlated insulator you want to try also DFT+U as an option, by the way. From your ouput I read that you are getting an AFM solution half spins up half down, with no net moment the reason being that elk finds this as the lowest groundstate for your starting configuration. If you are interested in something with a ferrimagnetic net moment I rather suggest you to put 5 spins up and 3 down in your input card which should result into something with a net...
Dear Andrea, first in CeO2 we have the case of an f0 element, means no electrons in the f-states. Thus the U-correction will shift the empty levels upwards, which is also the case from your DOS. What would be interesting to see would be to have a look at the f-states only (partial dos) to see what are really f-states in your plot but I would assume its the double peak. Is this correct? Regarding the effect of U it should make no difference to other codes. The only technical swtich you can do is to...
Dear Andrew, the eigenvectors are given in cartesian ! secondly the components are given in sqrt(amu) Bohr so to (amu=atomic mass unit) convert this vector into a atomic displacement you have to divide each component of the eigenvector with the respective sqrt of atomic mass it belongs to. So lets say you have NiO (2 atom unit cell) than the first column has to devide by sqrt(28) and the second by the respective weight of oxygen sqrt(8). You will need to do some scripting to create this superimposed...
Dear Marjana, sorry I missed your statement. Question 1 is your cell big and how are the forces after the third structural iteration? I somehow remember that lattice relaxation requieres somehow forces below a treshold but I may be wrong on this. best regards Michael
I think "latvopt" has to be set to either 1/2 to perform constrained or unconstrained structural relaxation. the default is 0, which does no lattice optimization. best regards michael
I think "latvopt" has to be set to either 1/2 to perform constrained or unconstrained structural relaxation. the default is 0, which does no lattice optimization. best regards michael
you can find a nice intro to that here https://arxiv.org/abs/1202.1831 and the basic idea is to compute the shift of the wannier centers for displacing each atom, which corresponds to the created electric polarization. happy computing Michael
Dear Brice, rgkmax 12 is a huge cutoff, so I guess you are checking convergence so first I would suggest check from rgkmax --> 6 - 8 how things change but remember check energy differnces rather than a single total energy if this does not converge you may go further. If you reach 12 and it did not converge, yet. You will have to increase the complementary grid by gmaxvr (which is the cutoff for the potential). For large enough values this message should go away. happy computing Michael
Hi everybody, Currently I am doing some computations on an insulating AFM material but encounter the problem of an oscillating k-convergence. Meaning that If I compute the energy difference between two distinct AFM states call them AFM1 and AFM2, the difference Delta E=E(AFM1)-E(AFM2) follows some Sin(NKPT) beahviour and never reaches a specific value. Any Idea how to go about that in general? best regards Michael PS: I hardly remember but is there an issue to use autokpt+radkpt in combination with...
Dear Brice, currently not that I know. best Michael
Currently, elk does not have an inbuild functionality to compute ferroelectric polarization. But with a little bit of handwork you can use the w90 interface with elk to compute the wanniercenters which allow you to determine the born effective charges. best Michael
Dear Lizard, I would use a python two liner and confer http://www.cryst.ehu.es/cryst/get_kvec.html for the corresponding settings
The g-factor is not calculated by the code direclty, but you can compute L and S and hence we convert it into a gfactor as M=muB(L+2S)=muBg*S best michael
Dear Brice, the point is that the ''default'' potential are delivered with default linearization energies matching the radii with which they have been generated. Hence if you alter the radii strongly you may also have to change the linearization energies as it is now a different ! best regards Michael
Actually it is a ''true'' magnetic field and enters the kohn-sham equations by B\cdot\sigma do not confuse this ''initial'' field with the constraining fields, which are Lagarange parameters, if you perform computations for constrained magnetic moments. For all cases there should be examples provided with elk best Michael
This is true, but to do so you need to initalize with a non-colinear structure and elk will than switch to the corresponding mode. best Michael
Dear Lizard, probably you set the field during the restart on each atom to low. So if it is <=1E-6 elk will not recognize it and set it to zero instead, So there are two options either set epslat 1E-8 which lower the level of elk recognizing small number or use a larger field and than set reducebf 0.1 so the field is faded out quickly. I prefer the former but both should give the same result. best Michael
the magic flag is maxlatvstp which is by default 30.
--> Both is possible you may want to check the manual and further the provided examples --> within the elk folder! --> Still in INFO.OUT (you should concentrate your questions in one thread of the forum!) --> bfcmt is the local magnetic field applied on each muffin tin sphere. It breaks the symmetry and thus trigger spin-polarized solutions. In most cases small fields are enough but for some f-elements different field lead to different solutions due to the high multiplicity of the f multiplets. happy...
you can find them in info.out --> here an example of the last line of one of my recent computations where under moments you find the magnetic moment within the muffin tin radius for each site. Note that this is only the ''spin'' moment best regards and happy computing Michael Charges : core : 32.00000000 valence : 148.0000000 interstitial : 18.04738829 muffin-tins (core leakage) species : 1 (Fe) atom 1 : 24.55491801 ( 0.9755440900E-10) atom 2 : 24.55491801 ( 0.9755440900E-10) atom 3 : 24.55492510...
Dear Brice, by default, elk will not break symmetries within to structural relaxation! So in this respect you are completly save. Actually, it is more often the case that one is looking for a lower symmetry state, which to its end you will need to compute phonons and check for soft modes as a hint. best regards and good computing Michael
Dear Harry, no problem, would be nice if you report back which file worked :-). best regards Michael Last, one remark a great workshop for all the ''symmetry'' stuff is the bilbao crytallographic workshops as listed -->( http://www.cryst.ehu.es/cryst/workshops_and_schools.html ). They also teach you how to use their tools in all kind of ways. If you have time/founding it is definitly worth going.
what you have to do is lets take the example here (short paste from my current project). This is one symmetry operation of the crystal, where the spatial translation is a shift of the lattice (T) and the spatial rotation a rotation matrix (R). In this case the former is zero and the latter is a inversion and a mirror in principle. Crystal symmetry : 2 spatial translation : 0.000000000 0.000000000 0.000000000 spatial rotation : 0 -1 0 -1 0 0 0 0 -1 next we take a k-point of my lattice of the reduced...
Dear Harry, I think the procedure would be to take the symmetry reduced set and than apply all operations listed in the SYMCRYS.OUT. Once this is done you should end up with the full k-point set. I assume that some of the k-points may appear several times. best regards Michael
Dear Andrew, could you check a 2nd thing, could you offcenter the k-point grid by vkloff so add vkloff 0.25 0.125 0.625 and check again, in some cases that ''healed'' rough curves. Moreover, could you share your input thus I could have a test for my own ..... best Michael
Dear Andrew, Long time ago "Anton Kozhevnikov" explained to me a similiar problem. That ultimatly somehow increasing RGKMAX leads to unstable situation. The conclusion of our discussion was the following maybe helpful for your situation as well: RGKmax control the plane wave coefficients outside of the MT in the interstitial, wheras lmaxapw controls the l-cutoff inside the MT. By construction of the basis one has to match one with the other. If you now increase RGKmax you somehow end up with a situation,...
Dear Robert, I had a similiar problem recently but with an added module. Although, I think elk is quite save in this respect. An idea for solving could you see how much memory your job takes from the job manager log files ? This could give you an idea where you are. Another possiblity is to remove all atoms from your cell only leaving one and see how it goes. Once this is running add more until the stage it crashes. This definitly would give a hint if the size of the task is a problem. best Mich...
Dear Robert, it would suprise me, but did you try to run without dmaxbnd? best Michael
Dear Robert, I think it is enough to specify Eu several times to get elk into a state treating these atoms as different. Exemplary here a modified version of the NiO LDA+U example, where I split the Ni sites although they access the same "Ni.in" file, they will be treated as individuals by the code. atoms 3 : nspecies 'Ni.in' : spfname 1 : natoms; atpos, bfcmt below 0.0 0.0 0.0 0.0 0.0 -0.01 'Ni.in' : spfname 1 : natoms; atpos, bfcmt below 0.5 0.5 0.5 0.0 0.0 0.01 'O.in' : spfname 2 : natoms; atpos,...
Dear Robert, how do you submit the job ? openMP ? MPI ? Sometimes I have the feeling openMP is more resource hungry. Anyway, also your system is big, which brings itself elk to the edge. best Michael
Dear Robert, okay than we can work on this could you try a strip down version of your elk.in and see if the message still persists. Note the demaxbnd at the end, which I added to increase search radius for the linearization energies. tasks 0 mixtype 1 spinpol .true. bfieldc 1.0 0.0 0.0 reducebf 0.75 xctype 3 ! is this on N or Fe lda+u 1 1 2 2 0.18 0.037 ngridk 4 4 1 demaxbnd 3.5
Dear Elio, In general you define the fcc lattice as avec 0.5 0.5 0.0 0.5 0.0 0.5 0.0 0.5 0.5 scale 7.646==a which is equivalent to the setting given in the example just one uses avec 1 1 0 1 0 1 0 1 1 scale 3.8267==a/2 So both settings are the same, which you use is a matter of taste. You can check GEOMETRY.OUT for confirmation :-) if you like. Back to your problem, I have run the computation further, and I have the feeling still something is not well here. For example, with "autolinengy=.true."...
I would suggest first to try a simplified version of the make file. Take out the MKL stuff and compile with ifort thus you should go with MAKE = make F90 = mpif90 F90_OPTS = -O3 -ip -unroll -no-prec-div -qopenmp F77 = mpif90 F77_OPTS = -O3 -ip -unroll -no-prec-div -qopenmp -assume protect_parens AR = ar LIB_SYS =. LAPACK and BLAS libraries LIB_LPK = lapack.a blas.a LIB_FFT = fftlib.a SRC_OMP = SRC_libxc = libxcifc_stub.f90 SRC_FFT = zfftifc.f90 does this run without crashing than ? best Michael
Dear Robert, some questions first: -- are you sure about the lattice constants are in bohrradius? -- I noticed you want to utilize lda+u and you want to correct the l=2 states at the N sites? is this correct? I would rather thought to apply +u to Fe? best Michael
could you post your makefile and give more insights in your elk.in ?
Dear Elio, To Fix your computation you can add first demaxbnd 3.5 this will increase the search radius for the linearization energy and the message will be gone about this problem next add lradstp 2 which increases the radial meshes and hence the message with the charge density is gone. Still adding this the system behaves strangely. Are youre sure about the lattice constants ? I found for PdS2 a=5.502 Ang == 10.3973 bohrradius. The units of lattice constants in elk are bohr ! Checking your structure...
Dear Harry, I guess this is a detail in the nitty-gritty of the APW+lo, which only either Lars or Kay could answer in detail. Just as a request could you check how different the value you obtain with autolinenegy (see LINENGY.OUT) is from the defaul of 0.15. best regards Michael
Dear Andrew, I used this feature several times and I assumed the implementation is done similiar to other codes, where to compensate the violation of the charge neutraltiy some background for the ewald summation is added. But I guess you find the detail in the code. Adding the charge will indeed change the energy particular the band energy as you have now more or less electrons. However, as the particle number changes it is not really possible to compare total energies of computations with differnet...
Ultrafast Relaxation Dynamics of the Antiferrodistortive Phase in Ca Doped SrTiO3 M. Porer, M. Fechner, E. M. Bothschafter, L. Rettig, M. Savoini, V. Esposito, J. Rittmann, M. Kubli, M. J. Neugebauer, E. Abreu, T. Kubacka, T. Huber, G. Lantz, S. Parchenko, S. Grübel, A. Paarmann, J. Noack, P. Beaud, G. Ingold, U. Aschauer, S. L. Johnson, and U. Staub, Phys. Rev. Lett. 121, 055701 (2018)
**Ultrafast Relaxation Dynamics of the Antiferrodistortive Phase in Ca Doped SrTiO3 M. Porer, M. Fechner, E. M. Bothschafter, L. Rettig, M. Savoini, V. Esposito, J. Rittmann, M. Kubli, M. J. Neugebauer, E. Abreu, T. Kubacka, T. Huber, G. Lantz, S. Parchenko, S. Grübel, A. Paarmann, J. Noack, P. Beaud, G. Ingold, U. Aschauer, S. L. Johnson, and U. Staub, Phys. Rev. Lett. 121, 055701 (2018) (https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.121.055701)
Dear Harry, finally I came up with the best solution for this problem. Actually, I was thinking already some time about this issue with identifying the characters of each PDOS entry, Finally, the best way is to also have a look on the eigenvalues of the lmirep, which displays the character. They are already computed and I just outputed them in a file "ELMIREP_VECTORS.OUT" (see replaced routine dos.f90 for elk-5.2.14). If you run now the code and compute the dos (task=10) elk will write the vectors...
Dear Harry, find here my patch for elk 4.0.15 (i did this some time ago) for the dos written out in real harmonics. To allow also for the old projections I converted lmirep to a integer and you can now select either lmirep new == old 0 == .false. 1 == .true. 2 == > did not exist now projects on real spherical harmonics Please note that I use for the proejctions the definitions according to https://en.wikipedia.org/wiki/Table_of_spherical_harmonics#Real_spherical_harmonics and also the sorting of...
Dear Harry, find here my patch for elk 4.0.15 (i did this some time ago) for the dos written out in real harmonics. To allow also for the old projections I converted lmirep to a integer and you can now select either lmirep new == old 0 == .false. 1 == .true. 2 == > did not exist now projects on real spherical harmonics Please note that I use for the proejctions the definitions according to https://en.wikipedia.org/wiki/Table_of_spherical_harmonics#Real_spherical_harmonics and also the sorting of...
Dear Zhangjin, the latter avec(:,:)=avec_(:,:)+deltast*strain(:,:,istrain) is correct as strain(:,:,istrain) already contains the lattice vectors. thus \eta_i=ei*A, whith \eta_i is the i'th strain tensor. if you have a look into genstrain.f90 you see how it is constructed best Michael
Dear Harry, in principle elk give you the real/tesseral spherical harmonic projections, the problem is to identify what is what. By default the lmirep=.true. tag will enforce that elk performs a symmetrization of the pdos of each atomic site according to its site symmetry. So for example in a octaheadron coordinate site such as the B site of an ABO3 perovskite or bcc iron the d-states will split in t2g and eg. If you check for such a system ELMIREP.OUT file you will find that in the l=2 block for...
Dear Saeed, as stated in the manual, if you want to constrain a direction you still have to give values here as correspond to the direction you want to constrain. So in full detail The fields you provide in the atoms block are the initial ''kick'' to induce a spin-polarization. The fields provided here are finally faded out controlled by the parmeter ''reducebf'' if you want to constrain now the system to a specific setting you use mommtfix and specifiy the configuration you want (remember the minus...
Dear Saeed, the reason is in the atoms block you define an intial field for the magnetic moments, whereas in the mommtfix you define the moment. as you can test if you specifiy only the field during a selfconsitent run the magnetic moments will develop in opposite direction to the initial field according V=m.B. If you next want to constrain these moments you should choose a ''configuration'' close to this state and hence define moments with opposite signs to the field. Otherwise, you calculation...
Okay, could you post the default output to check wether elk already computed the strain ? Anyway if you are trying to relax a surface, as I guess this from you input card. You may want to keep the lattice vectors fixed! The reason for this is that you should have enough vacuum to allow the atoms to relax in c-direction. Number two: what do you mean the atoms are spread inside the unit cell. From your previous post (elk.in and geometry.out) 1.) the elk shift was obvious and the atoms relaxed a bit...
Dear Shivam, this looks like a ''really'' big cell. Did you try a very low (rgkmax=6) and did you try to increase gmaxvr already. Moreover, what is the standard out for this computation ? best Michael
Dear Debarati, without seeing your ''elk.in'' it is hard to assist in such cases, if you put more details I will have a look. best Michael
Dear Anonymous, you may check your structure: your last three iron atoms and the first three Sn atoms occupy the same positions. This is not really a healthy situation. You may use the feature to put out a xrysden file and have a look at the structure to check. best Michael
Dear Eleni, by default elk shifts the entire lattice that the first atom (Mo in your case) exhibit 0 0 0 coordinates, which I guess is the source of your observation. You can suppress this by setting tshift .false. or just ignore it and go on with the shifted coordinates. best regards Michael
Dear Jammi, regarding your question 1) The term ''wrong'' sound more dramatic as it is. If you have a look on the isolated atomic filling of Ce (https://www.ptable.com/#Orbital) you find a single 5d1 electron and a single 4f1 one, whereas the confirguration elk uses to start with is 5d0 and 4f2. That different in occupation is what I was refereing to. In the end it shouldn't matter for a regular calculation but if you want to put something into the core you should be taking care. regarding two but...
Dear Mohammad, could you post INFO.OUT and the LOGS of standard out. Moreover, what is the favorable result ? best regards Michael
Dear SHIVAM, the simplest apporach to describe doping is either: 1) to use a large supercell and replace one atom with a dopant of your choice 2) second option is to manipulate the electron count to ''simulate'' electron or hole doping using chgexs (see manuael) Regarding your comment on Cu I guess if it doesent become metallic in the end a spin polarized state is not a stable minimum, You can post your input here and I could have a look. best Michael
Dear Joel, this will be a tricky case. Anyway we shouls start with the lin. energies warnings the rest may goes for free. 1.) I would put back the 1s Carbon state to core, since most of my computations never converge with a 1s state in valence. Could you provide some INFO.OUT of this computation and furthermore the LINENGY.OUT file to check which linearisation energies he does not find! 2.) you may set ''demaxbnd'' in the input card, which is the search radius for the linearisation energies to a...
Hi Shivam, I assume your question is more basi: As you already know in plot1d you specify number of points you want to go through and the number of plotting points inbetween. Finally, you list all the points according their position in recpiprocal space. For all space groups you find a list of the high-symmetry points here --> http://www.cryst.ehu.es/cryst/get_kvec.html and an example for AlAs in the elk folder @ examples/basic/Pb with space group Fm-3m (Number 225 according international tables)...
Dear Jammi, This calculation is indeed a tricky case and convergence is not straightforward to obtain. From the structure, you send me I failed to converge here anything with the frozen core method. Okay then I had a look at the chemistry, and we have from the periodic table the following configuration for Ce 6s2 4f1 5d1. Reflecting this configuration with the actual settings in elk, I realized that we are missing 5d1 to create an f1 core configuration. Hence I introduced the 5d1 and lowered the...
Hi Cenna, Just you should get sure you get an ordered magnetic moment by checking INFO.OUT. Regarding the readadu, you do not obtain speed by reading in the alpha value since its calculations takes only seconds. Instead, reading in alpha gives you more control since you always have the same value. This might be helpful for example if you are interested how properties alter with the structure and thus freeze in alpha. However, in setting alpha by yourself, you also choose in principle a specific kind...
okay, alpha=0 means AMF and alpha=1 means FLL having 0.23 I assume its more AMF. However, do you have a ordered spin moment? Here, we have to be careful now about the magnetic state. Dependening if you have a PM solution or real AFM state the alpha value might be quite different. Hence, you should deside which system you want to describe more the AFM insulator or the PM metal just you should check. From my experience in cuprates sometimes is difficult to get both stable solutions. best Michael PS:...
Dear Cenna, your input is a little bit counteracting. You are applying a local magnetic field on Cu to enforce a AFM state and at the same time you also apply a global magnetic field, which favors a FM state. From my experience I would suggest you to do the followin: 1) remove bfieldc 2) increase the field at the atoms to a value between 1 - 5 3) set reducebf = 0.8 -- 0.9 4) set mixing to linear this should speed up things a little bit. Actually, broyden mixing (for me in transitionmetal oxides)...
Dear Top Gun, the units in ''all'' dispersion relations are 1/bohr (x-coordinate) and energy/frequency is always given in Hartrees (y-coordinate) best regards Michael
Dear Top Gun, the units in ''all'' dispersion relations are 1/bohr (x-coordinate) and energy/frequency is always given in Hartrees (y-coordinate) best regards Michael
dear elio, yes you should since otherwise you create a contradictory situation. In principle it shouldn't matter to much since bfieldc is reduced anyway each iteration, however, one never knows. One more comment please remember that the interaction energy of of magnetic field and moment is E= m.B hence you should initialze both with opposite sign. this momfix 0.0 0.0 1 bfieldc 0.0 0.0 -1 note the minus for the field. It may not matter in your current calculation but if you ever do transitionmetaloxides...
Hi, Could you tell me which ifort you are using ? Did you try to compile other stuff/codes with you ifort installation does it work? Are there some strange messages during compilation ? Which operting system you are using? Another check could you run the Fe calculation within the magnetism folder in the examples and tell me if it go through? What are the magnetic moments you get ? best regards Michael
Dear Alain, you should check first to compile elk with the provided libraries for blas and lapack. I remember some incompatiblity with MKL libraries. I am using (intel/mkl/64/2017/2.174) which work but first check the provided once and then go on. best regards michael
Dear Elio, so first regarding the magnons: I think the principle idea of magnon calculations is similiar to phonons. This I conclude from the Ni example you find in "examples/TDDFT-magnetic-response/Ni-magnon/elk.in". You have to perform two calculations first : with vqlss = q and theta 0 second : with vqlss = q and theta 15 the energy dE than gives w(q) = (4/M) dE / sin^2(theta). Regarding the formula it should be still valid for alloys. The only problem I would assume is the alloy description by...
Dear Elie, if you are only interested in J_1^AA ...--> you don't really need to consider the spin-spiral stuff instead you may just take energy differences as a first approximations. Consequently, you write your Heisenberg Hamiltonian such as H= J1 S1.S2 + C, and then cycle through different magnetic arrangements S1=1 S2=1 & S1=1 S2=-1 to resolve C and J. Please note that here C contains all other Energy contributions. A more sophistacted approach is also given in Ref. [1] ( H. J. Xiang, E. J. Kan,...
Dear surinder, core files are often created when the job crashes and current state of the memory is written into this file. So I think its a problem with your cluster rather than with the code. Another issue could be the compliation, maybe. best regards Michael
Dear Anastaisa, this a MPI variant for make.inc to compile elk but I strongly suggest you to consult the administrators of your cluster to have a look on it. Moreover, they also could provide you with an example input script. best regards Michael MKLROOT = /cluster/intel/mkl MAKE = make F90 = mpif90 F90_OPTS = -O3 -ip -openmp -I$(MKLROOT)/include F77 = mpif90 F77_OPTS = -O3 -ip -openmp AR = ar LIB_SYS = LIB_LPK = -Wl,--start-group $(MKLROOT)/lib/intel64/libmkl_intel_lp64.a $(MKLROOT)/lib/intel64/libmkl_core.a...
Dear os1, I think this is possible but without stating more information on your calculation and what you consider litle, it is hard to say. best regards Michael
Dear JT, Firstly, this statement is true for all first-principle codes. The main problem is that the convergence of the total energy concerning all truncating parameters for the basis (rgkmax, lmaxapw ...) or sampling parameters for the total energy integration (k-points) is slow. Thus to achieve convergence of the value of total energy in the range of 10^{-6} eV you will need sizeable numerical setting. In contrast energy differences between two physical situations, i.e. an antiferromagnetic and...
Dear Chris, I think the problem is that if you go from a primitive to a larger unit-cell you obtain in some cases more symmetries. This behaviour appears often in in distorted perovskites, where some people prefer to use a pseudo-cubic unit cell instead of the smaller rhombhohedral. Anyway approaching you case I would suggest to simply follow the advise and change maxsymcrys within modmain.f90. Its value is set to 192 but just try to increase it to 256 and see what happens after recompilation. best...
In principle, I would be also interested in this topic but did you try to use tefvit .true. to switch to an iterative solver for the first variation equation. This speeded up some of my large cell calculations best Michael
Dear Jimmi, could you post the references to these papers? I would be very interested in this. In the meantime, the regular version of ELK has no DFPT for the spin-polarized system which consequently means no DFPT for computations with spin-orbit. The message you received from the code "Error(phonon): spin-polarised phonons not yet available." is a reflection on that. best regards Michael
Dear Harry, Wich states you get depends on your settings for core and the valence states. DMATMT contains all f-character of the valence electrons on the site. If you have 4f and 5f in the valence, you should get both. However, so far I am aware off, most of the species files put the 4f into the core once you're reached at least the 5f row in the periodic table. best regards Michael
1) If you do not specify oxygen, it will be constrained to 0! Thus the code implies that the moment should vanish. 2) If you don't wont to do that you should set oxygen constrained moment to 2000. One remark on this, once you set the moment the default elk version might complain that the symmetry is not correct. The problem arises because elk checks if your constrained moment is compatible with the B-field settings and stops the computation if it's not. The solution to that is to do one iteration...
Dear Cenna: Your statement is correct. One more advice, depending what system you are dealing with you might also switch off constraints on magnetically unrelated atoms. So for example in ABO_3 perovskites, you may want to constrain the magnetic ions only and leave oxygen unconstrained. This approach also speeds up the calculation in several cases! To do so, you can set the constrained moments on the "other" atoms to a value above 2000. Then the code will not constrain these atoms. Alternatively,...
Dear Cenna, in principle, mommtfix should have the opposite signs of the magnetic field applied to each atom because otherwise, the constraint will try to ''reverse'' each magnetic moment where it is applied. In the case you post here, this happens in the 2nd or 3rd iteration as you can see by checking INFO.OUT. Consequently, it somehow works, however, in other cases setting this tag like this will result in a long calculation with nearly non-convergence. Hence, I would recommend you to set ''always''...
In principle reducing MT does not need compensation. you may want to check the core leakage in INFO.OUT if its still small! otherwise I see no issue. best Michael