Hi everyone, I would just like to follow up on the discussions regarding magnetism in Titanium. I haven't examined this system, but you can also observe false magnetism in another metal, palladium. GGA pseudopotentials usually overestimate the experimental lattice constant and this slightly expanded lattice can lead to overestimates of the magnetic moment for the system. You may want to try the LDA pseudopotential for comparison. For phonon dispersions, the results you get from LDA and GGA are usually comparable. The smaller lattice constant due to LDA will push the phonon frequencies slightly higher, but I suspect they will still be in good agreement with experiment.
Palladium sits at the edge of magnetism and it is particularly sensitive to the pseudopotential you choose. Titanium may have similar problems. For my project, I was looking at potential magnetism on the nanoscale so it was important that my predicted bulk systems were non-magnetic! For some plane-wave calculations I did with Quantum Espresso, I found I needed a LDA pseudopotential with a semi-core d state and nonlinear corrections to get the proper bulk non-magnetic case (see J. Appl. Phys. 101, 09D503 (2007)). This problem with the GGA pseudopotential for Pd has also been observed with Siesta (see the Comment by S.S. Alexandre et al. Phys. Rev. Lett., 96, 079701 (2006)). This has also been observed by some people I talked to who used an FLAPW approach, although I don't think this was ever published. So my suggestion would be to try LDA and see if you still have this problem. If LDA works, it should be sufficient for your phonon dispersions. Best regards, Derek -- Derek Stewart, Ph. D. Scientific Computation Associate http://sites.google.com/site/dft4nano/ 250 Duffield Hall Cornell Nanoscale Facility (CNF) Ithaca, NY 14853 (607) 255-2856 > Dear Zhen, Salvador: > sorry if I missed something important in your previous exchange. > Anyway: > If you calculate regular metallic bulk Titanium and get it magnetic, > the cure to the problem is NOT to switch spin off, but try to find out > WHY it converges magnetic. Normally it should converge to > a nonmagnetic solution whatever your start spin configuration. > It is NOT like two metastable states, one magnetic and the other not. > In my opinion, it can be (apart from unaccomplished convergence) > only due to too large lattice constant, or too short basis functions > (sorry I could not find check your input/output files to check). > To get a better idea, calculate the density of state of your hcp Ti > and see if it resembles anything good. I bet you'll see several > narrow peaks instead. > Please do this before going on to phonon calculations. > Best regards > > Andrei Postnikov > >> >> Dear Zhen, yes you are right. >> You have already run your calculations, and get a large spin >> polarization. >> >> Now run another one, with the two flags indicated. If turning on those >> flags gives you lower energy, then the spin-polarized result is a >> metastable ground state, as Marcos indicated. In that case, always keep >> these flags and move on. >> >> Best regards, >> -Salvador. >> >> ----- Original Message ----- >> From: "Zhen Huang" <[email protected]> >> To: [email protected] >> Sent: Wednesday, March 3, 2010 10:39:58 AM GMT -05:00 US/Canada Eastern >> Subject: Re: [SIESTA-L] Fwd: Titanium Phonon Calculation Re: Titanium >> calculation help posted earlier >> >> Dear Both- >> >> A little bit confused. So Salvador you are saying I can run two tests. >> one >> w/o totalspin and fix spin one w/. If the one w/ totalspin 0.0 and >> fixspin >> true give lower total energy, I can move on to phonon calculation? >> Thanks >> >> Best Regards >> ------- >> Zhen (Alex) Huang >> Ph.D. Student >> Nanoscale Transport Research Group >> Laboratory for Computational Methods in Emerging Technologies >> Cooling Technologies Research Center >> School of Mechanical Engineering >> Purdue University >> Tel: 765 237 9733 >> >> >> >> On Wed, Mar 3, 2010 at 10:17 AM, Barraza-lopez, Salvador < >> [email protected] > wrote: >> >> >> >> Dear Zheng and Marcos, >> To check Marcos' point one could add the following lines to the input >> file: >> >> totalspin 0.0 >> fixspin True >> >> If the total energy with with these added lines is bigger (in absolute >> value) than the one you have just obtained, then you are fine, and just >> need to use this constraint all the time. >> >> >> Best regards, >> -Salvador. >> >> ----- Original Message ----- >> >> From: "Marcos VerÃssimo Alves" < [email protected] > >> To: [email protected] >> >> >> >> Sent: Wednesday, March 3, 2010 9:43:32 AM GMT -05:00 US/Canada Eastern >> Subject: Re: [SIESTA-L] Fwd: Titanium Phonon Calculation Re: Titanium >> calculation help posted earlier >> >> Salvador and Zhen, >> >> Forgive me if I'm wrong, but couldn't magnetic states be displayed in >> DFT calculations even if the material is not magnetic in its ground >> states? This could simply be a metastable ground state, and in this >> case what one should really compare is the total energy of the system >> in both states, magnetic and paramagnetic. What would really worry me >> is if, metallic Ti being paramagnetic, the magnetic state were lower >> in energy. >> >> And, on a personal note, I'd trust more the Abinit pseudo than the one >> generated in a plug-and-play fashion by the Octopus website - there is >> always a tendency to use the rc's supplied by default in the >> interface, which are not necessarily good ones. Moreover, the Octopus >> website interface to the pseudopotential generator allows for less >> flexibility than a mastery of the ATOM program file format. Abinit >> pseudos are often tried and tested, even though it is imperative to >> test every pseudo in simple cases before going for the real thing. >> Just my 10 cents on the subject. >> >> Cheers, >> >> Marcos >> >> On Wed, Mar 3, 2010 at 3:21 PM, Barraza-lopez, Salvador >> < [email protected] > wrote: >>> >>> Dear Zhen, >>> I would recommend that you change the cuttof radii in your pseudo and >>> regenerate it (perhaps 2.0 radii for all your channels, but you can >>> play >>> with it). Then test it again for polarization in the bulk. >>> If you never get spin polarization to be zero, try turning on >>> relativistic >>> corrections, and even core corrections. Just be methodic and go for >>> each >>> of >>> these things (radii, relativistic corrections, core corrections) >>> sequentially. The atom manual should be sufficient to guide you in the >>> process. The process should be: Generate your pseudo with atom, run a >>> SIESTA >>> calculation with that one and record the magnetization with a standard >>> basis >>> set, if polarization is not zero, go back and generate a new pseudo, >>> and >>> iterate as needed. >>> >>> Whenever your polarization is zero, you'll discover that the lattice >>> constant in the bulk should be reasonable; then you can proceed to >>> study >>> phonon dispersions. >>> >>> Best regards, >>> -Salvador. >>> >>> >>> >>> ----- Original Message ----- >>> From: "Zhen Huang" < [email protected] > >>> To: [email protected] >>> Sent: Tuesday, March 2, 2010 8:20:42 PM GMT -05:00 US/Canada Eastern >>> Subject: Re: [SIESTA-L] Fwd: Titanium Phonon Calculation Re: Titanium >>> calculation help posted earlier >>> >>> Dear Salvador- >>> I turned on the spin-polarized. I find the Qup-Qdown is not zero. >>> siesta: Total spin polarization (Qup-Qdown) = 4.856526. I am using GGA >>> PBE generated from http://www.tddft.org/programs/octopus/pseudo.php >>> Lodger >>> radii 3. I also tried the GGA psf converted from ABINIT which I >>> download >>> from SIESTA website the Qup-Qdown is about 4.939. Can you let me know >>> how >>> can I fix this? Thanks. >>> Best Regards >>> ------- >>> Zhen (Alex) Huang >>> Ph.D. Student >>> Nanoscale Transport Research Group >>> Laboratory for Computational Methods in Emerging Technologies >>> Cooling Technologies Research Center >>> School of Mechanical Engineering >>> Purdue University >>> Tel: 765 237 9733 >>> >>> >>> On Mon, Mar 1, 2010 at 10:14 AM, Barraza-lopez, Salvador >>> < [email protected] > wrote: >>>> >>>> Hi all, >>>> >>>> >>>> >>>> I've been following this post and want to mention something else that >>>> has >>>> not come out yet. Ti has in the valence two d and two s electrons. >>>> d-electrons are responsible for magnetic behavior. So, I'd like to >>>> ask: >>>> Have >>>> you taken your Ti pseudo (is it LDA, GGA pseudo?) and ran a >>>> spin-polarized(!) bulk calculation on the HCP phase (two atoms in unit >>>> cell) >>>> ? Don't worry about the basis now, just use the standard one. >>>> >>>> >>>> >>>> If you've done this, is your total spin up minus spin down is equal to >>>> zero? >>>> >>>> Titanium should be non-magnetic. If you get a large spin polarization >>>> here, there is no point in keep going further. >>>> >>>> If you are sure your bulk Ti run is non-magnetic, please disregard my >>>> post. >>>> >>>> >>>> >>>> Best regards, >>>> >>>> -Salvador. >>>> >>>> ----- Original Message ----- >>>> From: "Zhen Huang" < [email protected] > >>>> To: "SiestaMailingList" < [email protected] > >>>> Sent: Monday, March 1, 2010 8:19:00 AM GMT -05:00 US/Canada Eastern >>>> Subject: [SIESTA-L] Fwd: Titanium Phonon Calculation Re: Titanium >>>> calculation help posted earlier >>>> >>>> Dear Friends- >>>> I had trouble of reply my old post. This is a following question about >>>> Titanium. After I increased the kgrid_cutoff to 25 Ang. The results do >>>> become better. However, when I am running 3*3*3 supercell calculation >>>> the >>>> dispersion curve becomes like attached. I am running a 5Ã5*5 case (on >>>> 4 >>>> processors) now however, the calculation becomes really slow. I wonder >>>> what >>>> is the best way to run it faster. Again my input files are attached. >>>> Please >>>> give me some advice. Further, shall I comment kgrid_cutoff in >>>> vibrational >>>> properties calculation? I tried it but it gives me totally wrong >>>> results. I >>>> wonder if I can speed up the calculation by doing anything. I deeply >>>> appreciate your help. >>>> >>>> Best Regards >>>> ------- >>>> Zhen (Alex) Huang >>>> Ph.D. Student >>>> Nanoscale Transport Research Group >>>> Laboratory for Computational Methods in Emerging Technologies >>>> Cooling Technologies Research Center >>>> School of Mechanical Engineering >>>> Purdue University >>>> Tel: 765 237 9733 >>>> >>>> >>>> >>>> -- >>>> Salvador Barraza-Lopez >>>> Postdoctoral Fellow >>>> School of Physics >>>> The Georgia Institute of Technology >>>> >>>> Office N205 >>>> 837 State Street Atlanta, Georgia 30332-0430 U.S.A >>>> Tel: (404) 894-0892 Fax: (404) 894-9958 >> > >
