Activity for Andrew Shyichuk
-
Andrew Shyichuk posted a comment on discussion Elk Users
Dear Jerry, I've skipped the hr part for now. Other than that yes, it does work.
-
Dear Kay, Got the code, works wonders. Thanks!
-
Dear users, I am using this old Wannier90 thread for a new question. Namely, elk does not produce UNK00001.1 file and that makes plotting with Wannier90 impossible. Is there a reason for that, and how can this be implemented? Thank you
-
Dear Developers, I've noticed that the xlwin=[xlwin,str] line in readinput.f90 does not work, at least with some compilers. The produced array of lines contains gibberish and the line read last that ends up in the first element. My compiler is Intel 2022.1.0 and Elk is 9.2.12. My fix is a primitive one that works :) I created the additional xxlwin in modw90.f90, and in readinput.f90 the respective case reads: case('xlwin','wannierExtra') if (allocated(xlwin)) deallocate(xlwin) if (allocated(xxlwin))...
-
Congrats, and welcome to the "fiddling around with compilation" club.
-
Dear Swati, You've posted in an unrelated topic, but oh well. Warnings at the first few loops are no problem as long as they go away later. The problem shows up at a magnetic stage indicating that RMT's are likely not the issue. Metals are not really my thing, but'd say - try looking into +U approaches or an MGGA functional (RPP works well with elk). Both are quite some rabbit holes to fall into, but you might be having a strongly correlated system on hand, and those require special treatment. Look...
-
Alrighty, solved. I used ptnucl=f to make the nuclei non-point - as I suspected the DFT functional doing oddball things around the nuclei. Removing this option (which is internally set to t) solved the problem.
-
Dear Users, I found a peculiar case related to my adventure with phosphates. After some struggle with convergence issues of a mixed Sr-Y phosphate (energy suddenly jumping up after 5-10 loops), I've kind of solved it by making phosphorus the first specie, which resulted in base offset to P (center of PO4 group moved to the origin). This solved the jumping energy completely, and uncovered the subject of this post. The energy converged to about 1e-6 while the RMS-potential kept oscillating below 1e-3....
-
Thank you, that clears it.
-
Hi, With latvopt =/= 0 you will get your cell optimized and the strain will go away (that is, if task i 2 or 3). If you have to optimize - do only the the atomic positions with latvopt = 0. Strain will come from an overall cell size changes that can be induced e.g. by scale keyword. Also, as this is likely a slab - you shoud use partial scale and not total scale, see the manual.
-
Hi, With latvopt =/= 0 you will get your cell optimized and the strain will go away (that is, if task i 2 or 3). If you have to optimize - do only the the atomic positions with latvopt = 0. Strain will come from an overall cell size changes that can be induced e.g. by scale keyword.
-
Dear Users, I am running a project with YPO4. With default parameters (including RMTs and LOs) the calculated charge error was about 0.18e out of total (correct) 172e - which is a lot. So, I went to the forum, read the posts on RMTs of Si, found no posts on phosphates, messed around with LO's, lo's, LAPW, number of points, rgkmax, gmaxvr, none of which helped noteworthy. Finally, I optimized the RMTs in a primitive steepest-descent style and got an improvement: the error got reduced to about 0.008,...
-
Dear Kazu-s, Long story short - you gotta optimize RMT, rgkmax, number of k-points, etc. RMT first, lmaxo/lmaxi/lmaxapw later. What I am doing now is trying to minimize error in calculated charge. Sad story of trial and effort.
-
Dear Users, I calculated Epsilon_11 with RPA, TDDDFT and BSE for YPO4. The first two are qualitatively the same and only differ in intensities. BSE, on the other hand, is distinctly different. RPA peaks start at about the band gap edge, and BSE peaks are lower in energies - in line with exciton bands being slightly below band-to-band excitations. The noob question: does BSE give a complete optical absorption, or just the excitonic part? If its just excitonic - can I sum the two spectra, i.e. are...
-
How do I use this approach with a defect? In a regualr periodic calculation, if I add one defect per cell, I get, in principle, infinitely many defects in a periodic lattice. Is it possible to use the method to calculate a defect in a cell surrounded by defect-free cells?
-
How does the method scale in terms of CPU time? If I have an SCF with 27 k-points and task 700 with 27x27 = 729 Q-points, do I expect a x27 factor to the SCF loop time?
-
What happens if I specify avecu not corresponding to an integer number of unit cells? What if it has a rotation in respect to the unit cell?
-
How do I choose kappa-point grid size? I.e. the ArXiv paper says: * Ideally, the smallest distance between k-points should be greater than the largest distance between κ-points, i.e. |κ − κ′| < |k − k′|. This will ensure that the set k + κ does not overlap for any two k-points, which may lead to double counting and an over-complete basis set.* I'm not quite familiar with thinking in terms of reciprocal cell distances, especially when it comes to supercells. But, after some trial and error I got this:...
-
Dear Users, Ultra long-range calculations have been implemented in Elk. The respective papers are here and here. This is a superb method, and I am happy it was implemented. How about we make this post sticky and have the questions as separate comments?
-
Dear Elio, If you can - ask the admins for a newer compiler. If not - compile the newer compiler from scratch and keep it, as the code will need libraries from it. The same goes for OpenMP. Good luck! Andrew
-
Dear Farhad, From what I know - you have the right impression: EOS is done manually. As for geometry optimization - it's a hard topic for Elk. For me it did not work back at version 6.8.2: position optimization was fine, but with lattice optimization it got weird (and time consuming). In particular, Elk minimizes forces, which in my case corresponded to an increase in energy - which I find unphysical. On the other hand - I do not know much about geometry optimization algorithms and it's not for me...
-
Dear Vasiliy, With gmaxvr this high, use msmooth = 4. This kind of charge density error might indicate a too small basis (low rgkmax) or an incomplete calculation. Check the LINENGY.OUT file for any variable values (local orbital energies) the same as in the input (you can usually spot them by many trailing zeroes). This does not include the fixed value (APW linearization energy) which is 0.15 by default. Stderr will also contain linengy warnings if this is the case. Best regards. Andrew
-
Dear Farhad, Set maxscl to 0, or to 1 - the latter will result in 1 SCF loop. Best regards. Andrew
-
Dear Swarup, This is an anime forum a forum for Elk FP-LAPW quantum physics/chemistry code, not ELK analytics. Best regards. Andrew Context for the joke here.
-
Dear Kay, Thank you for clearing that out. What about scaling? How does RDMFT calculation scale with system size? Or, at least, what is the approximate ratio between SCF and RDMFT CPU time? Best regards. Andrii
-
Dear Kay, I've made a test on system with a 10.4 Angstrom side lengh and a 108x108x108 g-vector grid. The integral as a function of the NxNxN 3D plot grid is: N Integrated magnetization 20 5.5786380189537965 27 2.6195550437733353 40 1.540586033524166 54 1.207590381642519 70 1.0877206831044655 108 1.0073521869283015 135 0.9908741650541215 The best result is obtained without the interpolation (when N_plot = N_g_vector), and the value is what it should be. The bug was in my integration code: there was...
-
Dear Users, I've achieved my goal of the field-free magnetization. The answer was to reduce swidth. I have the following questions now: 1. Can I interpret my odd-electron non-magnetized system (majority spin density ~= minority spin density) as a paramagnetic state? After all, it's not just one cell with just one electron - the cell represents an infinite continuum of cells. A half of the "unpaired" electrons can be up, and another half can be down. Am I wrong? 2. Lets say I make a series of calculations...
-
Dear Kay, As I understand, maxthd1 will limit any first-level parallelization, the k-point loop or not. Is there maybe a keyword to specifically limit k-point parallelization only? I am trying to run RDMFT calculation and I get the impression that maxthd1 also limits density matrix (KMAT.OUT) calculation. Best regards. Andrew
-
Dear Users, The RDMFT paper by Sharma et al gives RDMFT power functional power parameter α of 0.7 or 0.65 (0.7 looks better from the paper). Another paper gives the value of 0.578. The RDMFT CECAM presentation gives the value of 0.565 and says, "Value of α is fixed to 0.565 and all calculations for extended systems are performed with this FIXED value. The theory is ab-initio in this sense." The previous slide shows the power functional, while the graph with the quoted text shows the Müller functional,...
-
Dear Users, The RDMFT paper by Sharma et al gives RDMFT power functional power parameter α of 0.7 or 0.65 (0.7 looks better from the paper). Another paper gives the value of 0.578. The RDMFT CECAM presentation gives the value of 0.565 and says, "Value of α is fixed to 0.565 and all calculations for extended systems are performed with this FIXED value. The theory is ab-initio in this sense." The previous slide shows the power functional, while the graph with the quoted text shows the Müller functional,...
-
Dear Developers, What should a 3D magnetization plot (task 73) sum up to? If I have the " total moment" per cell about 1 - I expect the 3D magnetization plot to integrate to (roughly) the same value. And, I know that for Elk 3D files, the integral over cell equals S * V / N, where S is the sum of values of the 3D file, V is cell volume and N is the number of points in the 3D file. Or, which is the same, the integral is the sum times the voxel volume. For my 3D files, I get the value of about 0.15,...
-
Dear Kay, There wasn't, it's a new PMAT.OUT. Come to think of it, it's header should contain size. Incomplete file would than result in an unxpected EOF error, right? Best regards. Andrew
-
Dear Users, I wonder if Elk will complain if task 121 was run with incomplete PMAT.OUT from an interrupted task 120. Does successful termination of task 121 indicates a complete PMAT.OUT? Thank you. Andrew
-
Dear Users, I'm trying to run a regularized SCAN calculation on a hexagonal lutetium oxide, input below. The problem is: it fails, no matter what I do. With r2 SCAN (xctype 100 497 498) it usually happens in 4 steps, with rSCAN (xctype 100 493 494) it can last, but then it fails. I tried running RPP mGGA first (xctype 100 209 9), and then use that STATE.OUT to start the rSCAN calculations. It helps them to make more steps, but eventually they fail. The fail is a rapid increase in the RMS, super high...
-
Dear Robert, I was able to converge your system in 34 scf loops with the input below. Soft Broyden mixing was the key, meaning fsmtype != 0 (which forces adaptive linear mixing) was not ok for this system. With the small muffin-tin control fileds, it happily converges to the correct magnetizations. Best regards. Andrew tasks 0 gmaxvr 17 rgkmax 8.0 msmooth 4 isgkmax -2 lradstp 1 xctype 3 broydpm 0.2 0.075 spinpol t spinorb f dft+u 1 1 1 3 0.293995 0.000000 ngridk 9 9 1 nwrite 1 avec 6.923937625 0.000000000...
-
UPD. The described method works, although not perfectly. I will stick to manual restarts for not.
-
Dear Robert, With fsmtype != 0 you don't need bfieldc and reducebf, that is for sure. I don't know if the code knows to ignore them in the said case. You likely do not need autoswidth here, it's an insulator, it does not matter much, but it is a spare degree of freedom. Add lradstp = 2 or 1. The calculation will get slower, but it might help, given the rapidly changing density of the 4f orbitals. Now, highq sets these: rgkmax=max(rgkmax,8.d0) gmaxvr=max(gmaxvr,16.d0) msmooth=max(msmooth,4) lmaxapw=max(lmaxapw,9)...
-
Dear Andrey, I did a quick grep through the source code, and bfinite does not show up there. It is an outdated information, apparently. The manual says: "In cases where the magnetic field is finite (for example when computing magnetic response) the external B-field energy reported in INFO.OUT should be added to the total by hand." So, there's that. Best regards. Andrew
-
Dear Robert, That depends on the system's convergence character. To me, FSM calculation started from a converged non-magnetized system (task=1) is more likely to converge properly, it's a safer approach. Moreover, it can be helpful if an FSM calculation from scratch does not work. You already have some result, I'd start with that. Best regards. Andrew
-
Dear Yuan, The warning is your answer. I assume your calculations converge to a wrong state where some linengies are not found. If it is indeed so - the results are plane wrong and unusable. See e.g. this thread, but also search this forum for the warning. It has been discussed in many details before. Best regards. Andrew
-
Dear Robert, 6.89 is fine, these are the moments inside muffin-tins only. 4f orbitals are compact, but not 100% confined to the radius of ~2 bohr of whatever the RMTs of Eu are in this case. I'd go with fixed moment calulations, see fsmtype keyword. Also, I'd bet that spin-orbit couplig should be important is this case. A good trick is to use cmagz=t to keep the system collinear (memory and time consumption will be almost the same as those of regular spin-polarized calculation). If you want to optimize...
-
Dear Users, I am trying to achive a high-spin state in a calculation without the control filed. I assume the field might introduce artifacts and thus I am checking out other options. For the sake of comparison, I do that on a geometry which has been optimized with the field (with high spin), and in which the magnetization has a distinct localization spot. The number of unpaired electrons is 1. So far, I've got an oscillating moment situation - as the calculation reaches convergence, the moment goes...
-
Dear Andrea, Try increasing the k-point grid and nempty. Also, consider using meta-GGA (e.g. xctype 100 209 12), although I am not sure if that will work with +U. I shoud do fine without +U though. Still, you are interested in super high energies, which are in principle not something you would analize with DFT. Consider CASPT2 or CCSDT methods in OpenMolcas with embedding AIMPs. The embedding data are likely to be available somewhere. Best regards. Andrew
-
Dear Youzhao, Default is APW, nxoapwlo=0. LAPW is nxoapwlo=1. But, given the fixed linengy local orbitals in the species files. the default would have been APW+l.o. in WIEN2k notation. With semi-core states, there are also variable-linengy local orbitals, which are named LO or L.O. in WIEN2k notation, and are present in many (most?) of the default species. Thus, the default is APW+l.o.+LO or APW+l.o.+L.O. Yeah. I hope I remeber that right, but just in case - search this forum for "APW+l.o.". Best...
-
Dear Robert, Linengy warnings are no problem only if they go away, eventually. At convergence, all linengies must be found. Also, muffint-tin charges must be roughly the same at the same sites. Consider the attached scripts that do the checks. I wonder why do you have this many nempty, default value should do fine. With this settings your calculation will converge to a ferromagnetic state. I do not know If that is what you want. For antiferromagnetic or other states, you'd have to use atomic moments...
-
Dear Youzhao, I guess it was, I did some plots with Elk 6.8.4 and they were fine. You can test it easily by calculating the plot of interest and then checing the values inside a 3D file. Also, you can integrate the file and see if the integral is about 2 for spin-unpolarized or about 1 for spin-polarized calculation. Best regards. Andrew
-
Dear Zhiwei, It can be tricky, especially with DFT+U. You do not need bfieldc. To break the symmetry the atomic fields are enough and can be even smaller. You can also try fsmtype for a desired atomic magnetizations. However, note that the target magnetization values apply to muffin-tin part only, which are smaller than the total magnetization related to a given atom. Reducebf 0.9 seems too high, the field gets down too fast. It'd go with atomic fields of 0.0001, no reducebf, and then a restart without...
-
Dear Users, I'm trying to calculate Fermi surface for a cubic cell with a~=20 bohr, 3x3x3 k-point grid, no reduction (low-ish symmetry). Task 100 (or 101) runs for 3 hours, and then fails with: Operating system error: cannot allocate memory. Allocation would exceed memory limit. This is not a kind of error that happens when a taks asks for too much memory. Moreover, it does not use more than 2 in this failed run. It looks like it requests more memory than the system has - which is strange, given...
-
Dear Developers, Elk relies heavily on FFT, and calls external FFT libraries. Inside the library, the FFT machine probably calculates the DFT matrices on every call. Please correct me if I'm wrong on this one. Elk calls FFT and IFFT tens of thousands times per SCF run, which (as far as I understand) are all of the same size - the g-vector grid size. And, in all of them, DFT matrices are caluclated. The question: if all of the above is correct, would it be more optimal to calculate the matrices once,...
-
Dear Developers, Elk relies heavily on FFT, and calls external FFT libraries. Inside the library, the FFT machine probably calculates the DFT matrices on every call. Please correct me if I'm wrong on this one. Elk calls FFT and IFFT tens of thousands times per SCF run, which (as far as I understand) are all of the same size - the g-vector grid size. And, in all of them, DFT matrices are caluclated. The question: if all of the above is correct, would it be more optimal to calculate the matrices once,...
-
Dear Gang, I see two things here. First, 1D/2D/3D plots correspond to geometry reported in GEOMETRY.OUT and INFO.OUT. By default, Elk shifts the atoms so that the first one of the least abundant specie (i.e. a dopant) is at the origin. The shifts are reported in the stdout (.o) files. I.e. when specifying geometry in elk.in, you must specify the plot lines for the geometry in INFO.OUT. Second thing - there is a discontinuity (albeit slight) at about 3. I'd try making EOS (energy-volume) curves in...
-
Would you please describe the fix you've made? Thanks. Andrew
-
Dear Ronald, How many lattice position optimization steps does your code make per one cell vector optimization step? Thank you, Andrew
-
Dear Users, I've noticed that, when optimizing atomic positions in elk, the energy goes down initially, but then it goes up. The forces, on the other hand, go down smoothly. I do lattice optimization too, but the thing happens in the initial cell size, while only atom positions are being optimized. No change in muffin-tin radii happens. The manual says, "The position of atom α is updated on step m of a geometry optimisation run using rm+1 =rm +τm (Fm +Fm−1 ), where τα is set to tau0atp for m = 0,...
-
When point nuclei are disabled (ptnucl = false) in a calculation with regularized SCAN MGGA (elk 6.8.4), an error occures: Warning(rdirac): maximum iterations exceeded Error(rdirac): zero wavefunction Input, e.g. in the GaAs MGGA example: ptnucl f xctype 100 493 494 Is this a proper behaviour? I thought, one of the reasons to not use point nuclei was to avoid instabilities in MGGA. Thank you. Andrew
-
Dear Kay, So, the effect is physical in principle, although we are talking crazy high fields? I am not going to make publishable conclusions based on that, at least not without an experiment. The question is rather: "There is this thing I can do with magnetic field in Elk, is it a legit reason to try an experiment?" I also noticed, that sometimes I cannot induce a stable magnetic moment with bfiledc = 1 or even more, but fsmtype = 1 results in a desired moment with a FSM global effective field of,...
-
Dear Users, It is known that magnetic fields are used in Elk in order to constrain or initiate magnetization. Is it possible to obtain the same effect in an actual material? For instance, can we force d electons from a low-spin ground state into a high-spin excited state by a static magnetic field? How strong the field should be? The only similar thing I found is Pauli paramagnetism, which is manifested as a shift between spin-up and spin-down bands under external magnetic fields. Are there other...
-
Dear Users and Developers, Is there a reason for Perdew-Zunger LDA being spin-unpolarized in Elk (xctype 2)? When trying a spin-polarized calculation, it results in an error: Error(init0): requested spin-polarised run with spin-unpolarised exchange-correlation functional Thank you. Andrew
-
Dear Jack, Would you share your Si input? Si is a troublesome element. Try searching this forum for tips on how to make it. The issue is in its 2s orbitals, which are safer to be treated as valence. For instance, a Si.in I currently use (taken from somewhere here, and maybe slightly changed): 'Si' : spsymb 'silicon' : spname -14.0000 : spzn 51196.73454 : spmass 0.534522E-06 1.5000 47.8169 400 : rminsp, rmt, rmaxsp, nrmt 7 : nstsp 1 0 1 2.00000 T : nsp, lsp, ksp, occsp, spcore 2 0 1 2.00000 F 2 1...
-
Elk output is in Hartree (energy) and Bohr (length). 1 Ha = 27.21138624599 eV 1 Bohr = 0.529177210904 Angstrom 1 Bohr^3 = 0.148184711473 Angstrom^3 In EOS plot, multiply Y by 27.21138624599 and X by 0.148184711473, and you got it.
-
Elk output is in Hartree (energy) and Bohr (length). 1 Ha = 27.21138624599 eV 1 Bohr = 0.529177210904 Angstrom With this, you can convert 1 Ha /B^3 into eV/A^3.
-
So, maybe a noob question. but. In DOS calculations, Elk reports spin quantization axis , which is set by sqados variable. The variable can be read from elk.in, which is not documented in the manual. The questions: - Does the axis even matter in a spin-unpolarized calculation? Would I get a different DOS with a non-z axis? - Does it make a physical sense to set the axis in a direction other than the converged magnetization direction? I.e. if I set it perpendicular to the magnetization, would I get...
-
Dear Miro, The fact that "Number of MPI processes :" is printed twice reporting one process, means that your MPI configuration is wrong. The run command is likely the culprit. See this post for more details. https://sourceforge.net/p/elk/discussion/897820/thread/bb25c53e70/ However, the system you run it at is defining. Supercomputer centers should provide manuals. Best regards. Andrew
-
Dear Yuri, Please check these topics: https://sourceforge.net/p/elk/discussion/897820/thread/bb25c53e70/ https://sourceforge.net/p/elk/discussion/897822/thread/fbc813ce12/ Good luck. Andrew
-
Dear Miro, I use my own scripts, that is why I do not know of the other codes. Bugs might occur from the input convention being missed or something. Your xyz has 307 in the header, whike it has 310 "atoms" if we count the vectors as atoms. For me, I changed that and named the vectors "Tv". Vector naming might matter. The cell is sane, it replicates flawlessly. Here's the file. But. You are going to have some trouble converging this. Vacuum alone will cost - a lot. I thus recommend ultrasoft pseudopotentials,...
-
Dear Miro, This: Info(checkmt): reduced muffin-tin radius of species 1 (Si) from 2.2000 to 0.0921 Info(checkmt): reduced muffin-tin radius of species 3 (H) from 1.4000 to 0.2511 Info(checkmt): reduced muffin-tin radius of species 4 (O) from 1.8000 to 0.0034 means that your geometry is very wrong. Your coordinates are cartesian, not fractional. Start with checking IADIST.OUT for the culprit, there must be some atoms there that are too close. Also, the third vector (0 0 50) looks like it was manually...
-
Dear Milo, This: Info(checkmt): reduced muffin-tin radius of species 1 (Si) from 2.2000 to 0.0921 Info(checkmt): reduced muffin-tin radius of species 3 (H) from 1.4000 to 0.2511 Info(checkmt): reduced muffin-tin radius of species 4 (O) from 1.8000 to 0.0034 means that your geometry is very wrong. Your coordinates are cartesian, not fractional. Start with checking IADIST.OUT for the culprit, there must be some atoms there that are too close. Also, the third vector (0 0 50) looks like it was manually...
-
Oxygen Vacancy, Oxygen Vacancy–Vacancy Pairs, and Frenkel Defects in Cubic Lutetium Oxide A. Shyichuk, E. Zych J. Phys. Chem. C 2020, 124, 28, 14945–14962 https://pubs.acs.org/doi/abs/10.1021/acs.jpcc.0c00974
-
Dear Cenna, "I did get the 0.15 minimum energy without autolinegy turned on. " - 0.15 is the default one and constant, this is expected. Bad optics was probably coming from a bad convergence. Regarding your questions. 1. Not sure. I usually search this forum, the matter was heavily discussed back at the day. In principle, the variable linearization energy must be roughly at Fermi level for semicore states, and below it for "less-semicore" states. If it's more than 0.5 Ha above - the LO is probably...
-
Dear Naman, From what I know, electrons in species files are the initial occupations. The final occupations are found self-consistently. Why would you define those from the start? For question 2, please refer to this topic and the manual. Next, atomic PDOS is actually a muffin-tin PDOS. I am almost sure that Fe 3d orbitals will be spatially larger than the muffin-tin sphere, and thus some portion of them will sit in the interstitial. You might try estimating the "total" occupancy from TDOS. Say,...
-
Dear Cenna, It took me some effort (maybe a hundred runs) and the whole set of "my typical" tricks (which mostly failed), but at the end I've had it covered. The zip contains the input files, elk.in in the final. Feel free to play around with the parameters, maybe it can be improved even more. Some notes and comments: - Increasing broydpm parameters results in a more aggressive mixing, reducing chances of convergence. - I've heard somewhere that odd-sized k-point grids are better, but that was for...
-
Dear Cenna, It took me some effort (maybe a hundred runs) and the whole set of "my typical" tricks (which mostly failed), but at the end I've had it covered. The zip contains the input files, elk.in in the final. Feel free to play around with the parameters, maybe it can be improved even more. Some notes and comments: - Increasing broydpm parameters results in a more aggressive mixing, reducing chances of convergence. - I've heard somewhere that odd-sized k-point grids are better, but that was for...
-
Dear Andrea, I am doing a related calculation right now, on Lu2O3. The problem with convergence is the first thing to solve. Having a stable SCF (task 0/1) is the basis for further calculations. I recommend trying the following: This setting generally helps in geometry-related jobs, just make sure that there is no atom at fractional coordinates 0.25 0.125 0.625. vkloff 0.25 0.125 0.625 Use high quality: highq t I prefer using manual MT radii. Check different values for Ce in the 2-2.5 range. Use...
-
Dear Karima, Maybe I was not clear. The series of calculations with different q is needed. But, that is not my area, and I cannot give you a more detailed answer. I'd recommend contacting Lars Nordström (https://katalog.uu.se/profile/?id=N96-772) in regard of this matter. Best regards. Andrew
-
Dear Kay, I did the same calculations will ultrasoft pseudopotentials as well, and got something like 98.5%. I cannot be 100% sure that the PAW result is correct. I did not make the PAW datasets and have no experience in making them from scratch - I cannot thus be sure that the ones I use are correct. The augmentation radii do not overlap, the code does not complain, it converges smoothly, EOS is smooth. Lu2O3 DOS from PAW (with 4f as valence) looks almost exactly the same as the DOS from Elk. It...
-
Dear Kay, Thank you. From your results, the minimum is at 95.1% of the initial / experimental cell size. I've got 98.1% in my best attempts. In the same calculation using PAW I get 99.5%. Thus, either 98% is the LDA-optimal cell size for my input, while some error compensation occurs in my PAW calculation, or we are still missing something here. Also, a small update - I tried bringing back the local orbital on oxygen (present in the default specie, missing in the one I mainly use). The resulting...
-
Dear Karima, A filed of 0.01 will break the symmetry. If the system "wants" to be high-spin, it would get high-spin. A field of 4.0 is quite large, and can whack your system (well, potentially) - i.e. force it into a non-physical spin state. As for spin spirals - see the manual for vqlss, ssdph, and spinsprl. If you want to get an energy dependence on spin-spiral propagation vector direction - you'd need to make a series of calculations with different vqlss. And I think no field is needed. Best regards....
-
Dear Kay, Did you happen to get any improvements on my input? In a meantime, I tried two options, highq and spin-orbit coupling. Both calculations got me the same minimum position. What else can it be? Thank you. Andrew.
-
Dear Karima, What do you mean by "large" or "small"? To start a calculation without a field just set the fields to zero. However, if you want the field to grow during the calculation - you may try reducebf value larger than 1. Note that reducebf changes the field in geometric progression, i.e. B(i+1) = B(i)*reducebf. Best regards. Andrew
-
Dear Karima, Reducebf option gradually reduces the field on every iteration. The purpose is: use larger field to break symmetry, but then, gradually let the system to find the minimum without the field. The field becoming negligible is exactly the purpose of this keyword. That, however, only works when the moment is stable without the field. That is not always the case. It happens that reducing the field also reduces the moment. Sometimes the system gets unphysical - zero total moment with odd number...
-
Dear Karima, Reducebf option gradually reduces the field on every iteration. The purpose is: use larger field to break symmetry, but then, gradually let the system to find the minimum without the field. The field becoming negligible is exactly the purpose of this keyword. That, however, only works when the moment is stable without the field. That is not always the case. It happens that reducing the field also reduces the moment. Sometimes the system gets unphysical - zero total moment with odd number...
-
Dear Karima, There is no such feature in Elk. You'd have to do it manually. Best regards. Andrew
-
Dear Karima, There is no such feature in Elk. You'd have to do it manually. Best regards. Andrew
-
Dear Jack, This is a fine input, although gmaxvr 25 seems to be quite a lot. I'd use 15 and no msmooth. Worth trying lradstp = 2 or 1 (even more points for MT grid). I'd try EOS next. Best regards. Andrew
-
Dear Kay, Using primitive cell was indeed a good idea. I does not affect the minimum position, but it runs quite faster and is thus a better test subject. Best regards. Andrew
-
Dear Jack, From my experience, elk geometry optimization is very tricky. I see that your P3_121 cell vectors have length of 5.65, 5.65, 4.77 Angstrom, while here they are 4.51, 4.51, 5.96 A. Yours seem to be way off. First things to check here are: 1. Muffin-tin radii. Make sure they do not change: compare radii in Te.in and INFO.OUT. Changing RMTs result in changing basis, making geometry optimization weird. I'd say you need at least 10% of bond length between the MTs (i.e. RMT = 0.45 x R(Te-Te))....
-
Dear Y.Gu, From what I know, elk executable does not handle MPI submission. I.e. it should be executed as "mpirun -nc 2 elk elk.in" other than just ./elk elk.in. I recommend checking this post for more. Thus, you should probably do something like: srun -N 2 -n 2 -c 24 mpirun -nc 2 ~/elk-6.3.2/src/elk elk.in That, however, depends on how to properly submit multinode / MPI jobs on that particular server, mpirun is just one possible option. Also, just to be sure, check if your 1-node task does indeed...
-
Dear Users, Some time ago I had a problem with cubic Lu2O3 EOS, which was rough. The roughness problem has been fixed thanks to Michael Fechner and vkloff = 0.25 0.125 0.625 trick. Now, I have another issue. With JTH PAW pseudopotentials (4f in valence, no +U), LDA and Quantum Espresso, I get energy minimum at 99.5% of the experimental cell size, which is great. With Elk and LDA, I get energy minimum at 98.1%, which is too small (even for LDA) and strange. The two caluclations have almost identical...
-
Dear Karima, Non-collinear magnetism is used when control magnetic fields are applied along axis other than z. Muffin-tin fileds are specified in elk.in, columns 4-6 of atom specification. Overall filed is specified by bfieldc. Best regards. Andrew
-
Dear Kay, In the zip, there are two series of inputs, with the respective outputs. Thank you. Andrew
-
Dear Users, Some time ago I had a problem with cubic Lu2O3 EOS, which was rough. The roughness problem has been fixed thanks to Michael Fechner and vkloff = 0.25 0.125 0.625 trick. Now, I have another issue. With JTH PAW pseudopotentials (4f in valence, no +U), LDA and Quantum Espresso, I get energy minimum at 99.5% of the experimental cell size, which is great. With Elk and LDA, I get energy minimum at 98.1%, which is too small (even for LDA) and strange. The two caluclations have almost identical...
-
Dear Users, Some time ago I had a problem with cubic Lu2O3 EOS, which was rough. The roughness problem has been fixed thanks to Michael Fechner and vkloff = 0.25 0.125 0.625 trick. Now, I have another issue. With JTH PAW pseudopotentials (4f in valence, no +U), LDA and Quantum Espresso, I get energy minimum at 99.5% of the experimental cell size, which is great. With Elk and LDA, I get energy minimum at 98.1%, which is too small (even for LDA) and strange. The two caluclations have almost identical...
-
Dear Users, I've made a short benchmark of different FFT routines on a Ryzen machine. I've used YBCO and CuFeS2 inputs from Elk basic examples, with increased k-point grids for a longer run time. Elk was compiled with GCC 9 with built-in libs, except for FFT, which was either: - built-in - precompiled FFTW 3.3.8, from repository (Fedora 30 Server) - manually compiled FFTW 3.3.8 (--enable-sse2 --enable-avx2 ) Workin directroies and scratch were located on a NVMe RAID 0 (stripped volume), and nwrite...
-
Dear Robert, Momfix works via self-consistent changes in the cell magnetic field (or atomic fileds) in order to approach the target moment. It is thus not exactly fixed. You do not need to use reducebf and bfieldc with momfix. Best regards. Andrew
-
Dear Developers, Currently, when potential-only mGGA is used, exchange and correlation energies are zero. I would like to calculate them with a functional of choice. What I expect to get is, for instance, LDA exchange and correlation energies from an mGGA density. Would that work if I converge an mGGA calculation, and then run an LDA calculaiton with maxscl=1? Thank you. Best regards. Andrew
-
Dear Kay, I'd like to add / check something. From the previous discussions, I remember that MKL calls happen at thread depth 4. Thus, having, say, 10-20 threads, even a threaded MKL will most likely be executed in a single-thread mode. Is it the same with MKL FFT threads? Also, with FFTW, Elk requires libfftw3.a, not libfftw3_omp.a and not libfftw3_threads.a. Is it thus true, that Elk controls threads in the FFT part and calls single-threaded FFTW? If so, is it also true for MKL FFT? Would that be...
-
Dear Developers, In Elk 6.3.2, lmaxmat is not used any more. What have changed since then? What can be used instead to achieve the same effect? Thanks. Andrew
-
Dear Pietro, I tried something like this back in the day with elk 6.2.8, using a sequential MKL. (I dimly remember having some troubles with threaded MKL, I guess) It worked with both openMP and MPI (although I did just a few tests). The important thing was that FFT routine has virtually no impact on Elk performance. I tried the built-in one, a manually compiled FFTW, a supercomputer FFTW, and MKL - there were some differences within the error threshold (i.e. running the same job gave something like...
-
I've missed this post and asked the same question. Consider this answer by Michael Fechner. Best regards. Andrew
-
Dear Michel, Thank you for the detailed answer. Best regards. Andrew
1 >