I tend to agree with Natalie's conclusions. Unfortunately I am not aware of any simple solution. All meta-GGA functionals do strange things in systems with regions of vanishing charge, and SCAN is the one that gives less problems (!). SCF divergence is maybe extreme, but there are also problems like this one: https://gitlab.com/QEF/q-e/issues/32, indicating that one may need a surprisingly dense FFT grid to get reliable results
Paolo On Wed, Nov 21, 2018 at 8:22 PM Holzwarth, Natalie <nata...@wfu.edu> wrote: > I have looked into the SCAN functional for an isolated atom, also finding > that Vxc increases to unphysically large values as the density decreases > exponentially. We were able to demonstrate for ourselves that the SCAN > functional itself is responsible for this behavior due to the appearance of > the density in denominators of various terms. It is my understanding > that the SCAN developers are aware of this issue and are working on a > remedy. In the meantime, perhaps it would be good to use the SCAN > functional only for systems that do not have exponentially decreasing > densities. At least that was my conclusion. Sincerely, Natalie > Holzwarth > > N. A. W. Holzwarth email: > nata...@wfu.edu > Department of Physics web: > http://www.wfu.edu/~natalie > Wake Forest University phone: > 1-336-758-5510 > Winston-Salem, NC 27109 USA office: Rm. 300 Olin > Physical Lab > > > On Wed, Nov 21, 2018 at 1:26 PM Filippo Savazzi <filo.sava...@gmail.com> > wrote: > >> Thanks Giuseppe, >> I’ll give it a try, I didn’t think about setting the occupation from >> input. >> Thank you for your help. >> >> Best, >> >> Filippo >> >> >> >> Filippo Savazzi, PhD Student >> Politecnico di Torino, Torino, Italy >> >> > On 21 Nov 2018, at 19:00, Giuseppe Mattioli < >> giuseppe.matti...@ism.cnr.it> wrote: >> > >> > >> > Dear Filippo >> > Isolated open-shell atoms are tricky. Please look here for hints >> > >> > yourpathtoQE6.3/PW/examples/example05 >> > >> > HTH >> > Giuseppe >> > >> > Quoting Filippo Savazzi <filo.sava...@gmail.com>: >> > >> >> Dear QE users, >> >> >> >> I’m Filippo Savazzi, a PhD student from Politecnico di Torino, Italy. >> >> I’m using QE since a couple of years, and just browsing the archives >> of this mailing list or enquiring the mighty google I’ve been lucky enough >> that this is actually the first time I have to post a question. >> >> Thanks everybody for the indirect support you gave me in these years. >> >> >> >> I’m calculating absorption energies of oxygen on a coronene molecule, >> as part of a little evaluation of different available XC functionals that >> I’m doing. I’m calculating the absorption energy both with respect to the >> molecular oxygen and with atomic oxygen. >> >> My problems arise when I try to calculate the energy of a single atom >> of oxygen using SCAN. In this case the energy diverges until the SCF fails. >> I have no troubles running the calculation on the actual system (coronene + >> bridged O), neither on a molecular oxygen (triplet O2); the issue is just >> localised on the single oxygen atom. My first guess was that the >> implementation of SCAN has either troubles in dealing with isolated systems >> or with spin polarization, but I ruled out these issues when I was able to >> calculate the single point energy of an oxygen molecule. Are you aware of >> any other problems using SCAN in these conditions? >> >> In the following I attach the input for pw.x I’m using, as well as a >> brief example of the SCF part in the output. I already played with cell >> dimension (thinking too much vacuum could have been the problem), as well >> as with ecutwfc and ecutrho (so indirectly with the integration grid), and >> then, just to give it a chance, with the diagonalization method and >> beta-mixing. >> >> >> >> Thank you in advance for your attention. >> >> >> >> Best, >> >> Filippo >> >> >> >> >> >> Filippo Savazzi, PhD Student >> >> Politecnico di Torino, Torino, Italy >> >> >> >> >> >> >> >> >> >> &CONTROL >> >> calculation = ’scf', >> >> restart_mode='from_scratch', >> >> pseudo_dir = '/home/filippo/pseudo', >> >> outdir='./tmp_o', >> >> prefix='oxy', >> >> / >> >> &SYSTEM >> >> ibrav=6, >> >> celldm(1)=20, >> >> celldm(3)=1, >> >> nat=1, >> >> ntyp=1, >> >> input_dft='scan', >> >> ecutwfc=100, >> >> tot_magnetization=2, >> >> nspin=2, >> >> / >> >> &ELECTRONS >> >> diagonalization = 'cg', >> >> electron_maxstep = 300, >> >> mixing_mode = 'local-TF', >> >> mixing_beta = 0.7, >> >> conv_thr = 1.0d-7, >> >> / >> >> &IONS >> >> ion_dynamics='damp', >> >> upscale=1000, >> >> / >> >> ATOMIC_SPECIES >> >> O 15.99 O_ONCV_PBE-1.0.upf >> >> ATOMIC_POSITIONS {angstrom} >> >> O 0.0000 0.0000 0.0000 >> >> K_POINTS{gamma} >> >> >> >> >> >> >> >> >> >> EXTRACT OF OUTPUT: >> >> >> >> Dense grid: 540376 G-vectors FFT dimensions: ( 128, 128, 128) >> >> >> >> Estimated max dynamical RAM per process > 116.06 MB >> >> >> >> Estimated total dynamical RAM > 1.36 GB >> >> Generating pointlists ... >> >> new r_m : 0.4125 (alat units) 8.2500 (a.u.) for type 1 >> >> >> >> Initial potential from superposition of free atoms >> >> >> >> starting charge 5.99905, renormalised to 6.00000 >> >> >> >> negative rho (up,down): 4.917E-05 4.917E-05 >> >> Starting wfcs are random >> >> >> >> total cpu time spent up to now is 3.1 secs >> >> >> >> Self-consistent Calculation >> >> >> >> iteration # 1 ecut= 100.00 Ry beta= 0.70 >> >> CG style diagonalization >> >> ethr = 1.00E-02, avg # of iterations = 9.6 >> >> >> >> negative rho (up,down): 1.044E-03 1.363E-03 >> >> >> >> total cpu time spent up to now is 240.5 secs >> >> >> >> total energy = 2429.63502350 Ry >> >> Harris-Foulkes estimate = -31.41616940 Ry >> >> estimated scf accuracy < 3.28037744 Ry >> >> >> >> total magnetization = 2.00 Bohr mag/cell >> >> absolute magnetization = 2.00 Bohr mag/cell >> >> >> >> iteration # 2 ecut= 100.00 Ry beta= 0.70 >> >> CG style diagonalization >> >> c_bands: 4 eigenvalues not converged >> >> ethr = 1.00E-02, avg # of iterations = 13.6 >> >> >> >> negative rho (up,down): 2.824E-03 3.815E-03 >> >> >> >> total cpu time spent up to now is 331.7 secs >> >> >> >> total energy = 33638.43350785 Ry >> >> Harris-Foulkes estimate = 318.68713998 Ry >> >> estimated scf accuracy < 0.37936007 Ry >> >> >> >> total magnetization = 2.00 Bohr mag/cell >> >> absolute magnetization = 2.00 Bohr mag/cell >> > >> > >> > >> > GIUSEPPE MATTIOLI >> > CNR - ISTITUTO DI STRUTTURA DELLA MATERIA >> > Via Salaria Km 29,300 - C.P. 10 >> > I-00015 - Monterotondo Scalo (RM) >> > Mob (*preferred*) +39 373 7305625 >> > Tel + 39 06 90672342 - Fax +39 06 90672316 >> > E-mail: <giuseppe.matti...@ism.cnr.it> >> > >> > _______________________________________________ >> > users mailing list >> > users@lists.quantum-espresso.org >> > https://lists.quantum-espresso.org/mailman/listinfo/users >> >> _______________________________________________ >> users mailing list >> users@lists.quantum-espresso.org >> https://lists.quantum-espresso.org/mailman/listinfo/users > > _______________________________________________ > users mailing list > users@lists.quantum-espresso.org > https://lists.quantum-espresso.org/mailman/listinfo/users -- Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222
_______________________________________________ users mailing list users@lists.quantum-espresso.org https://lists.quantum-espresso.org/mailman/listinfo/users