[Pw_forum] contradictory results of electron-phonon interaction for different k-grids

高淼 miaogao at ruc.edu.cn
Wed Dec 12 13:33:39 CET 2012 

 
> Dear all pw users:
> 
> I'm using QE-5.0.1 to calculate electron-phonon interaction for a ternary system (ibrav=4) with Li, B, and C.
> But I get contradictory results for different k-grids.
> 
> Calculation details:
> I choose ecutwfc =40.0, ecutrho=480.0, combined with smearing = 'gaussian', degauss = 0.02 Ry.
> The lattice parameters have been relaxed for different k-grid and pseudo potentials to get
> the corresponding equilibrium point. For example, the lattice parameters for pbe US potentials
> are about a=2.84 angstrom  and c=7.12 angstrom. I set conv_thr =  1.0d-14 and tr2_ph=1.0d-18
> in the scf and phonon calculation, respectively.
> 
> I use three different groups of k-grids to test the calculated results.
> the first group:
> densh  k-grid: 24x24x12
> coarse k-grid: 12x12x6
>            q-grid: 8x8x4
> the second group:
> densh  k-grid: 32x32x16
> coarse k-grid: 16x16x8
>            q-grid: 8x8x4
> the third group:
> densh  k-grid: 36x36x24
> coarse k-grid: 18x18x12
>            q-grid: 6x6x4
> 
> 
> Different pseudo potentials are also tested.
> 
> PBE:
> Li  6.941   Li.pbe-s-van_ak.UPF
> C   12.011  C.pbe-van_ak.UPF
> B   10.81   B.pbe-n-van_ak.UPF
> 
> PW91:
> Li  6.941   Li.pw91-s-van_ak.UPF
> C   12.011  C.pw91-van_ak.UPF
> B   10.81   B.pw91-n-van.UPF
> 
> PZ (combined NCPP and USPP):
> Li  6.941   Li.pz-s-van_ak.UPF
> C   12.011  C.pz-van_ak.UPF
> B   10.81   B.pz-bhs.UPF
> 
> PZ-vbc:
> Li  6.941   Li.pz-n-vbc.UPF
> C   12.011  C.pz-vbc.UPF
> B   10.81   B.pz-vbc.UPF
> 
> Input file for electron-phonon interaction:
> 
>  &inputph
>   tr2_ph=1.0d-18,
>   prefix='LiBC',
>   fildvscf='LiBCdv',
>   amass(1)=6.941,
>   amass(2)=12.011,
>   amass(3)=10.81
>   outdir='/home/gm/work/tmp/',
>   fildyn='LiBC.dyn',
>   electron_phonon='interpolated'
>   trans=.true.,
>   ldisp=.true.
>   nq1=6, nq2=6, nq3=4
> 
> My questions are as follows.
> 
> (1)The electron-phonon interaction parameter lambda is extremely large for the lowest phonon
> branch at some q points. For example for pbe-USPP, the largest lambda for the first group is 2.1725
> at q = (    0.000000000  -0.577350269   0.099769209 ), for second group is 19.4839 at
> q = (    0.125000000   0.360843918   0.000000000 ), and for third group is 3.2192 at
> q = (    0.000000000  -0.577350269   0.099718425 ). And the similar behaviors for PZ-NCPP.
> Is the extremely large lambda for lowest phonon branch reasonable?
> (2)for the third group k-grid with PBE-USPP, when asr=simple, there is a small imaginary
> frequency about -8 cm^-1 around Gamma to A (0,0,pi/c). But when asr=crystal, there is no
> imaginary frequency along the high symmetry lines. While for PZ-PP, when asr=simple,
> there is no imaginary frequency along the high symmetry lines. So which result should I trust?
> (3) For the second group, after applid asr, there are lots of imaginary frequencies, with the
> maximum about -167 cm^-1 at q=(0.186667  0.323316  0.199453). The result contradict that of
> the third group. But the first group k-grid show similar behavior with the third group,
> why does the second group k-grid show different results?
> 
> Could anyone give me some clue? I am greatly indebted to you for your help. Thanks again!
> 
> Miao Gao
> ------------------
> Ph.D. student of Department of Physics,
> Renmin University of China.
> ------------------
> 
> 
> 
> 
> 
> 
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: http://pwscf.org/pipermail/pw_forum/attachments/20121212/1fe2fb77/attachment-0001.html 
> 
> ------------------------------
> 
> Message: 4
> Date: Wed, 12 Dec 2012 09:57:31 +0100
> From: Lorenzo Paulatto <lorenzo.paulatto at impmc.upmc.fr>
> Subject: Re: [Pw_forum] contradictory results of electron-phonon
> 	interaction for different k-grids
> To: PWSCF Forum <pw_forum at pwscf.org>
> Message-ID:
> 	<CAG+GtJcLaqNdgYmLWnQiFQB5X2hnWv_XJNz-mEuhCLiSO4iEng at mail.gmail.com>
> Content-Type: text/plain; charset="utf-8"
> 
> On 12 December 2012 05:42, ?? <miaogao at ruc.edu.cn> wrote:
> 
> > (3) For the second group, after applid asr, there are lots of imaginary
> > frequencies, with the
> >
> > maximum about -167 cm^-1 at q=(0.186667  0.323316  0.199453). The result
> > contradict that of
> > the third group. But the first group k-grid show similar behavior with the
> > third group,
> > why does the second group k-grid show different results?
> >
> > Dear Miao Gao,
> the equilibrium lattice parameter and atomic positions depend on the choice
> of functional: you should perform an independent accurate variable-cell
> relax for each set of pseudopotentials.  By accurate I mean something like
>     forc_conv_thr=1.d-5
>     etot_conv_thr=1.d-6
> which in turn will require conv_thr of order 1.d-12 or tighter.
> 
> High-symmetry materials (i.e. when the atomic sites are Wyckoff positions
> with no free parameter) tend to always give "good" phonon dispersions, even
> if they are not properly relaxed; when the symmetry is lower even a small
> displacement from the equilibrium will give some negative frequency.
> 
> bests
> 
> 
> 
> 
> -- 
> Dr. Lorenzo Paulatto
> IdR @ IMPMC -- CNRS & Universit? Paris 6
> phone: +33 (0)1 44275 084 / skype: paulatz
> www:   http://www-int.impmc.upmc.fr/~paulatto/
> mail:  23-24/4?16 Bo?te courrier 115, 4 place Jussieu 75252 Paris C?dex 05
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: http://pwscf.org/pipermail/pw_forum/attachments/20121212/87c277df/attachment-0001.html 
> 
> ------------------------------
> 

Dear Dr. Lorenzo Paulatto:
Thanks for your quick reply. I have done the relaxation for each set
of pseudopotentials. For example the lattice parameters are:
a=2.841552688, c=7.123940931, for PBE-USPP
a=2.841820413, c=7.145618764, for PW91-USPP
a=2.787457537, c=6.762293607, for PZ-USPP

Just as you have mentioned, all the boron and carbon atoms occupied
the high symmetric Wyckoff positions, except for lithium. But I have
seriously relax the unit cell, with the total press until less than
0.02 kbar and the force acted on Li atoms about 1.d-5 (i.e. forc_conv_thr=1.d-5)
and etot_conv_thr=1.d-7. For example, the total energies of last three structures
in the relaxation are: 
!    total energy              =     -29.77186289 Ry
!    total energy              =     -29.77186291 Ry
!    total energy              =     -29.77186292 Ry. 

By the way, I can reproduce the results of MgB2 on the literatures.
The extremely large lambda for the lowest phonon branch never occur
in the calculation of MgB2, for all sets of pseudopotentials.
I remember Gosia had said (Malgorzata Wierzbowska <wierzbom at tcd.ie>) 
"US gives too big total lambda. The reason is because Eliashberg function,
in case of US, doesn't fall down to zero for small frequencies. Small
frequencies are for phonon vectors close to Gamma point."

But in my case, for second group with PBE-USPP, lambda is 19.4839 at
q = (0.125000000   0.360843918   0.000000000), which is not close to Gamma point.
And I have tested the PZ-NCPP, this situation become even worse. Any suggestion?

Best wishes

Miao Gao
------------------
Ph.D. student of Department of Physics, 
Renmin University of China.
------------------

More information about the users mailing list