[QE-users] Lambda ad Tc for Pb(111)

Paolo Giannozzi p.giannozzi at gmail.com
Mon Jun 15 15:06:09 CEST 2020


Look into lambda.f90 and figure out where and why it gives NaN. Most likely
there is a division by zero or a square root of a negative quantity.
Once upon the time, one needed to get rid of acoustic phonons at q=0 in
lambda.x in order to get sensible results. This was done by simply skipping
the first three phonons. I don't know how it works now (if it works)

Paolo

On Mon, Jun 15, 2020 at 2:53 PM Maciej Szary <maciej.szary at put.poznan.pl>
wrote:

> Dear Professor Giannozzi,
>
> sorry if I didn't specify. "NaN" referred explicitly to Tc calculations
> done by lambda.x. I do a pw.x calculation and I get a convergence, ph.x run
> is rather long, still it produces positive frequencies e.g.
>
>
> * ***************************************************************************
> *     freq (    1) =       0.819632 [THz] =      27.339964 [cm-1]*
> *     freq (    2) =       0.914078 [THz] =      30.490364 [cm-1]*
> *...*
> *     freq (   18) =       2.063295 [THz] =      68.824112 [cm-1]*
>
> * ***************************************************************************
>
> Next I use q2r.x. I do get some warnings e.g.:
>
> * Broadening =      0.005*
> *      q-space grid ok, #points =   36*
>
> *      fft-check warning: sum of imaginary terms = 3.497134E-06*
>
> but based on the previous forum posts, this seems acceptable.
>
>
> Subsequently I use matdyn.x and get lambda file containing:
>
> * Broadening   0.0050 lambda       1.7637 dos(Ef) 21.5379 omega_ln [K]
> 49.4388*
> * Broadening   0.0100 lambda       1.7028 dos(Ef) 21.2277 omega_ln [K]
> 49.7600*
> *...*
> * Broadening   0.0500 lambda       1.7499 dos(Ef) 22.1428 omega_ln [K]
> 48.1738*
>
>
> (shouldn't this be enough to calculate Tc using McMillan formula?)
>
>
> Finally I use lambda.x and get:
>
> *     lambda =      NaN (        NaN )  <log w>=      NaN K  N(Ef)=
> 21.537940 at degauss= 0.005*
> *     lambda =      NaN (        NaN )  <log w>=      NaN K  N(Ef)=
> 21.227733 at degauss= 0.010*
> *...*
> *     lambda =      NaN (        NaN )  <log w>=      NaN K  N(Ef)=
> 22.142765 at degauss= 0.050*
> *lambda        omega_log          T_c*
> *       NaN           NaN                NaN*
> *       NaN           NaN                NaN*
> *...*
> *       NaN           NaN                NaN*
>
>
> I've been redoing this procedure, with different parameters listed in my
> previous email.
>
> Regards,
> Maciej Szary
>
> --
> Research Assistant,
> Institute of Physics,
> Poznan University of Technology
> Piotrowo 3A, 61-138 Poznan
>
>
>
> On 6/14/20 3:50 PM, Maciej Szary wrote:
>
> Dear QE users,
>
> I'm trying to calculate electron-phonon coefficient and the critical
> superconducting temperature of Pb(111). As an introduction I've done the
> [PHonon/examples/example03] (bulk Al), and I've reproduced the results
> successfully. However, in case of Pb(111), I can't get Tc for the system. I
> only get "NaN". I understand that, this is often a result of imaginary
> components in the dynamical matrices, thus I've been redoing the
> calculations with different sets of parameters, however, to no avail.
>
> In scf I've been changing:
>
> 1. ecutwfc, and ecutrho (current values 80, and 650 Ry, respectively)
> 2. conv_thr (currently 1.0d-14)
> 3. k-point mesh (currently 36x36x1)
> 4. Pb pseudopotentials (currently Pb.rel-pbe-n-nc.UPF, previously
> Pb.rel-pbe-dn-rrkjus_psl.0.2.2.UPF)
> 5. Slab thickness (6-12 layers)
>
> In ph I've been testing:
>
> 1. tr2_ph (currently 1.0d-16, and up to 1.0d-17 with
> Pb.rel-pbe-dn-rrkjus_psl.0.2.2.UPF)
> 2. electron_phonon (interpolated, and simple)
> 3. q mesh (form 2x2x1 to 6x6x1)
>
> As a first step I've done the vc-relax, to relax the lattice constant of
> the slab. Next I perform relaxation with SOC included (etot_conv_thr, and
> forc_conv_thr= both 1.0d-6). Next I do the scf run e.g.:
>
> --------------------------------------------------------
> * &system*
> *    ibrav=  4, *
> *    a=3.503660834*
> *    c=40*
> *    nat= 6, ntyp= 1,*
> *    ecutwfc =80.0,*
> *    ecutrho = 650,    *
> *    occupations='smearing', smearing='mv', degauss=0.05*
> *    lspinorb=.true.,    noncolin=.true.,
> starting_spin_angle=.true.,     starting_magnetization=0.0,*
> *    la2F = .true.,*
> *    nbnd = 48 *
> */*
> * &electrons*
> *    mixing_mode = 'plain'*
> *    mixing_beta = 0.5*
> *    conv_thr =  1.0d-14*
> * /*
> * ATOMIC_SPECIES*
> * Pb  207.2  Pb.rel-pbe-n-nc.UPF*
> *ATOMIC_POSITIONS angstrom*
> *Pb       0.000000000   0.000000000  19.078230578*
> *Pb       0.000000000   0.000000000  27.596055131*
> *Pb       1.751830417   1.011419906  16.119173776*
> *Pb       1.751830417   1.011419906  24.874997245*
> *Pb       3.503661844   2.022839813  13.398232981*
> *Pb       3.503661844   2.022839813  21.915874849*
> *K_POINTS automatic*
> *36 36 1 0 0 0*
> --------------------------------------------------------
>
> this is followed by ph.x run:
>
> --------------------------------------------------------
> *&inputph*
> *    outdir='Files/',*
> *    prefix='QE'*
> *      fildvscf='aldv',*
> *  tr2_ph = 1.0d-16*
> *  amass(1) = 207.2*
> *  fildyn = 'Pb.dyn'*
> *  alpha_mix=0.2*
> *  electron_phonon='interpolated',*
> *  el_ph_sigma=0.005, *
> *  el_ph_nsigma=10,*
> *  trans=.true.,*
> *  ldisp=.true.*
> *  nq1 = 6, nq2 = 6, nq3 = 1*
> *  nogg = .true.*
> *  asr = .true.*
> */*
> --------------------------------------------------------
>
> Next are q2r.x,
>
>
> --------------------------------------------------------
> *&input*
> *  fildyn = 'Pb.dyn'*
> *  zasr = 'crystal'*
> *  flfrc = 'Pb.q661.fc'*
> *  la2F=.true.*
> */*
> --------------------------------------------------------
>
> matdyn.x,
>
>
> --------------------------------------------------------
> *&input*
> *    asr= 'simple'*
> *    flfrc = 'Pb.q661.fc', flfrq='Pb.q661.freq', la2F=.true., dos=.true.*
> *    fldos='phonon.dos', nk1=60, nk2=60, nk3=1, ndos=50*
> * /*
> --------------------------------------------------------
>
> and lambda.x
>
> --------------------------------------------------------
> *5  0.12  1    ! emax (something more than highest phonon mode in THz),
> degauss, smearing method *
> *    7*
> *    1   0.0000000  0.0000000  0.0000000   1.00*
> *    2   0.1666667  0.0962250  0.0000000   6.00*
> *    3   0.3333333  0.1924501  0.0000000   6.00*
> *    4   0.5000000  0.2886751  0.0000000   3.00*
> *    5   0.1666667  0.2886751  0.0000000   6.00*
> *    6   0.3333333  0.3849002  0.0000000  12.00*
> *    7   0.3333333  0.5773503  0.0000000   2.00*
> *elph_dir/elph.inp_lambda.1 *
> *elph_dir/elph.inp_lambda.2 *
> *elph_dir/elph.inp_lambda.3 *
> *elph_dir/elph.inp_lambda.4 *
> *elph_dir/elph.inp_lambda.5 *
> *elph_dir/elph.inp_lambda.6 *
> *elph_dir/elph.inp_lambda.7 *
> *0.10*
> --------------------------------------------------------
>
> The parameters I've used are relatively high in comparison to the Pb fcc
> example given by S. Poncé in 2018. Output of ph.x lacks negative
> frequencies, and DOS seems also fine:
>
> * # Frequency[cm^-1] DOS PDOS*
> *  0.0000000000E+00  0.0000000000E+00  0.0000E+00  0.0000E+00  0.0000E+00
> 0.0000E+00  0.0000E+00  0.0000E+00*
> *  1.4306338313E+00  3.8122112950E-03  6.3283E-04  6.3953E-04  6.3336E-04
> 6.3251E-04  6.4093E-04  6.3306E-04*
> *  2.8612676626E+00  6.7934053704E-03  1.1376E-03  1.1328E-03  1.1244E-03
> 1.1215E-03  1.1385E-03  1.1386E-03*
> *  4.2919014939E+00  9.5022844921E-03  1.6112E-03  1.5734E-03  1.5628E-03
> 1.5658E-03  1.5730E-03  1.6162E-03*
>
>
> *  5.7225353252E+00  1.2941007092E-02  2.2167E-03  2.1293E-03  2.1167E-03
> 2.1337E-03  2.1145E-03  2.2302E-03 **...*
>
> Also matdyn.x produces "lambda" file with:
>
> * Broadening   0.0050 lambda       1.7637 dos(Ef) 21.5379 omega_ln [K]
> 49.4388*
> * Broadening   0.0100 lambda       1.7028 dos(Ef) 21.2277 omega_ln [K]
> 49.7600*
> * Broadening   0.0150 lambda       1.6783 dos(Ef) 21.0885 omega_ln [K]
> 49.5935*
> * Broadening   0.0200 lambda       1.6881 dos(Ef) 21.1817 omega_ln [K]
> 49.4624*
> * Broadening   0.0250 lambda       1.7060 dos(Ef) 21.3799 omega_ln [K]
> 49.3380*
> * Broadening   0.0300 lambda       1.7218 dos(Ef) 21.5945 omega_ln [K]
> 49.1667*
> * Broadening   0.0350 lambda       1.7335 dos(Ef) 21.7888 omega_ln [K]
> 48.9500*
> * Broadening   0.0400 lambda       1.7414 dos(Ef) 21.9474 omega_ln [K]
> 48.7041*
> * Broadening   0.0450 lambda       1.7466 dos(Ef) 22.0647 omega_ln [K]
> 48.4423*
> * Broadening   0.0500 lambda       1.7499 dos(Ef) 22.1428 omega_ln [K]
> 48.1738*
>
> so shouldn't Tc be just a product of substitution into McMillan formula
> using values of omega_ln and lambda?
>
> Regards,
> Maciej Szary
>
> --
> Research Assistant,
> Institute of Physics,
> Poznan University of Technology
> Piotrowo 3A, 61-138 Poznan
>
>
> _______________________________________________
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
> users mailing list users at 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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20200615/0827c584/attachment.html>


More information about the users mailing list