Raman tensor

[eamouj78]

Sorry to bother you again. I got the Raman tensor at the end in this form for the 6 modes in graphene:

iel = 1
imode 1
-0.37159591E-001 -0.52084047E-001 0.13240895E+000 0.19372395E+000 -0.17649985E-002 -0.85285974E-003
0.13123341E+000 0.19603174E+000 -0.55708697E+000 -0.83045415E+000 0.19182040E-002 0.33954273E-002
-0.16223602E-002 -0.10028858E-002 0.15134916E-002 0.42465079E-002 -0.96255900E-003 0.73297461E-005
imode 2
-0.26585471E-001 -0.47464614E-001 0.12016822E+000 0.19654698E+000 0.16474102E-002 0.79528175E-003
0.12526254E+000 0.18780781E+000 -0.56933152E+000 -0.82210804E+000 -0.17942764E-002 -0.31708013E-002
0.15135202E-002 0.93664677E-003 -0.14130627E-002 -0.39612914E-002 -0.12384872E-002 0.85248738E-005
imode 3
-0.24341952E-001 -0.46925004E-001 0.11871116E+000 0.19395248E+000 0.21419466E-002 0.14369943E-002
0.11981237E+000 0.19211791E+000 -0.56292108E+000 -0.82651065E+000 -0.20289575E-002 -0.58334131E-002
0.21459822E-002 0.10812116E-002 -0.23379332E-002 -0.40514232E-002 -0.12694978E-002 0.12332381E-004
imode 4
0.39665632E-001 -0.38162862E-001 -0.17566636E+000 0.15218742E+000 0.40775402E-003 -0.23708537E-002
-0.17647463E+000 0.15372406E+000 0.75824689E+000 -0.65196752E+000 -0.70361579E-002 0.96142789E-002
-0.12348243E-002 -0.62126469E-003 0.13461721E-002 0.23303317E-002 -0.73364526E-003 0.70712627E-005
imode 5
-0.63188046E-001 -0.10047349E+000 -0.14041033E+000 0.40326534E-001 0.10032682E-002 0.90981463E-002
-0.15505757E+000 0.48039922E-001 0.99976989E+000 -0.21451481E-001 0.50860006E-001 -0.20388863E-001
0.90796428E-002 0.56169331E-002 -0.85023172E-002 -0.23744854E-001 -0.55617229E-003 0.61687156E-005
imode 6
0.89964247E-001 0.19874744E-001 -0.25488075E+000 -0.33511382E-001 -0.73833215E-006 0.11328906E-005
-0.27172025E+000 -0.24944500E-001 0.99666016E+000 0.81661001E-001 -0.51734809E-005 -0.81933241E-005
0.71338285E-006 0.22891055E-006 -0.12473664E-004 0.23328142E-005 -0.21004546E-002 0.16022080E-004

Let us focus on, say imode6.
1)Which represents the real part and the imaginary part of the tensor? How can we differentiate these?
2) How did you deduce that a and b (in the Raman tensors) are 0.1 and 1, respectively?

Thank you in advance

[nguyen-group]

Hi,

1. Let us take a set of values in imode 6 as (0.89964247E-001 0.19874744E-001), then 0.89964247E-001 is the real part, and 0.19874744E-001 is the imaginary part.
2. In the example of graphene with 60x60x1, our calculated results as:
   imode 5
   0.11774040E-001 0.67078001E-002 0.10368596E+000 0.99460619E+000 0.27814410E-004 -0.35555809E-005
   0.10368397E+000 0.99461029E+000 0.15074375E-002 -0.73996073E-002 -0.11586300E-005 -0.14580835E-004
   0.26534192E-004 0.74163811E-005 -0.56446092E-005 -0.43630915E-007 0.48308319E-003 -0.38809653E-005
   imode 6
   0.10460818E+000 0.99386902E+000 0.35508057E-002 -0.74211520E-002 -0.25745551E-005 -0.10840885E-005
   0.35575602E-002 -0.74204142E-002 -0.11627329E+000 -0.99321726E+000 0.13699722E-004 -0.91606077E-005
   -0.55019198E-005 -0.87527308E-005 0.28276571E-004 -0.18669124E-005 -0.28221599E-003 0.22620888E-005

You can see that a+ib = 0.10460818 + i0.99386902, so that is why a = 0.1 and b =0.99.
I note that a and b are very sensitive to the selected k-points. There is no rule for the convergence test of k-points for a and b. However, if the (a^2 + b^2) values of R_11 ~ R_22 and much higher than other Raman tensor components, as shown in bold text in imode 6, we can say it is a good choice.

In your case, you might use the different k-points or parameters compared with the example. We appreciate if you could do more tests with other k-points to check the value of a and b.

[eamouj78]

Once more, thank you very much for your time and patience.

The results I showed were a quick test to see if I can manage to obtain the Raman spectra using the raman.x executable. I will do more rigorous tests to investigate the convergence of "a" and "b" with denser k-points.

[eamouj78]

Dear all,

I have been performing some tests on the graphene Raman spectra. My results converged for 96x96x1. However I am getting completely different results in contradiction with the results reported in the example. The s+s+ values are very high, whereas they should be very low:

Here is a sample for a 72x72x1 grid :

My results:

#Raman spectra
#Laser energy 2.33 eV
#Raman shift (cm-1), Raman Spectra (XX,YY,ZZ,s+s+,s-s-,XY,YX,s+s-,s-s+)
1500.0000 0.28451797E+002 0.12499374E+004 0.10303849E+000 0.33009247E+003 0.31607869E+003 0.44149220E+002 0.42980392E+002 0.35006759E+003 0.36928010E+003
1500.1821 0.28623881E+002 0.12528658E+004 0.10358907E+000 0.33088534E+003 0.31677346E+003 0.44287541E+002 0.43115356E+002 0.35137037E+003 0.36986340E+003
1500.3643 0.28797674E+002 0.12558251E+004 0.10414516E+000 0.33168656E+003 0.31747562E+003 0.44427288E+002 0.43251711E+002 0.35268642E+003 0.37045316E+003
1500.5464 0.28973199E+002 0.12588157E+004 0.10470682E+000 0.33249626E+003 0.31818526E+003 0.44568479E+002 0.43389476E+002 0.35401592E+003 0.37104946E+003
1500.7286 0.29150478E+002 0.12618382E+004 0.10527413E+000 0.33331454E+003 0.31890250E+003 0.44711134E+002 0.43528669E+002 0.35535903E+003 0.37165239E+003
1500.9107 0.29329536E+002 0.12648928E+004 0.10584716E+000 0.33414151E+003 0.31962742E+003 0.44855273E+002 0.43669309E+002 0.35671594E+003 0.37226203E+003
1501.0929 0.29510396E+002 0.12679800E+004 0.10642599E+000 0.33497730E+003 0.32036014E+003 0.45000915E+002 0.43811415E+002 0.35808684E+003 0.37287849E+003

Results reported in code:

1500.0000 0.64057143E+003 0.63952548E+003 0.10761745E-003 0.28166636E-001 0.28180382E-001 0.64062004E+003 0.64062193E+003 0.12806222E+004 0.12806603E+004
1500.2506 0.64880016E+003 0.64774078E+003 0.10899987E-003 0.28528442E-001 0.28542365E-001 0.64884957E+003 0.64885148E+003 0.12970732E+004 0.12971118E+004
1500.5013 0.65718838E+003 0.65611530E+003 0.11040909E-003 0.28897261E-001 0.28911363E-001 0.65723861E+003 0.65724055E+003 0.13138430E+004 0.13138821E+004
1500.7519 0.66574024E+003 0.66465320E+003 0.11184580E-003 0.29273275E-001 0.29287560E-001 0.66579131E+003 0.66579328E+003 0.13309399E+004 0.13309795E+004
1501.0025 0.67446002E+003 0.67335873E+003 0.11331072E-003 0.29656672E-001 0.29671144E-001 0.67451195E+003 0.67451394E+003 0.13483726E+004 0.13484127E+004
1501.2531 0.68335214E+003 0.68223633E+003 0.11480459E-003 0.30047647E-001 0.30062309E-001 0.68340495E+003 0.68340696E+003 0.13661498E+004 0.13661905E+004
1501.5038 0.69242116E+003 0.69129055E+003 0.11632819E-003 0.30446401E-001 0.30461257E-001 0.69247488E+003 0.69247692E+003 0.13842807E+004 0.13843219E+004
1501.7544 0.70167182E+003 0.70052610E+003 0.11788230E-003 0.30853140E-001 0.30868194E-001 0.70172646E+003 0.70172853E+003 0.14027747E+004 0.14028165E+004
1502.0050 0.71110899E+003 0.70994786E+003 0.11946774E-003 0.31268081E-001 0.31283337E-001 0.71116458E+003 0.71116667E+003 0.14216416E+004 0.14216839E+004

What could have possibly gone wrong

[nguyen-group]

Hi,

I don't have any idea to see the output. Could you give me the input file?

[eamouj78]

My fault. I should have sent the input files. Since I could not attaching. I will copy-paste

Regards

scf.in:

&control
prefix='graphene'
calculation='scf',
tstress=.true.,
tprnfor=.true.,
pseudo_dir = '~/PP/'
wf_collect=.true.
outdir='/scratch/90022a/eamouj78/graphene'
wfcdir='/scratch/90022a/eamouj78/graphene'
etot_conv_thr=1.0D-6
forc_conv_thr=1.0D-5
/
&system
ibrav= 4, celldm(1) =4.6614811157162, celldm(3)=6.5, nat=2, ntyp= 1, ecutwfc = 70, ecutrho=700, degauss=0.007, nbnd=10, occupations='smearing', smearing='gaussian', assume_isolated='2D'

/
&electrons
conv_thr=1.D-8,
electron_maxstep=300
mixing_beta=0.1D0,
mixing_mode='plain'
/
&ions
ion_dynamics ='bfgs'
/
&cell
cell_dynamics='bfgs'
press=0.0
cell_dofree='fixc'
/
ATOMIC_SPECIES
C 12.0107 C.pbe-paw.upf

ATOMIC_POSITIONS crystal
C 0.000000000 -0.000000000 0.000000000
C 0.333333000 -0.333330000 0.000000000

K_POINTS automatic
72 72 1 0 0 0

bands_mat file:
&BANDS
prefix = 'graphene'
outdir='/scratch/90022a/eamouj78/graphene'
filband = 'graphene.bands'
lp = .true.
filp = 'graphene.dvec'
/

ph_mat file:

phonons
&inputph
prefix = 'graphene'
recover=.true.
outdir ='/scratch/90022a/eamouj78/graphene'
tr2_ph = 1.0d-18
fildyn = 'graphene.dyn'
fildvscf = 'dvscf'
electron_phonon = 'epc'
/
0.0 0.0 0.0

raman.in file:

&inputraman
prefix = 'graphene'
outdir='~/graphene/RAMAN/72x72x1/'
fil_dvec = 'graphene.dvec'
fil_elph = 'graphene.elph'
sorb = .false.
circular_pol =.true.
nonpol = .false.
plot_matele_opt =.false.
plot_matele_elph = .false.
plot_raman_k = .false.
gamma = 0.05
gamma_raman = 0.0002
rs_start = 1500.0
rs_end = 1600.0
nrs = 550
elaser1 = 2.33
/

[nguyen-group]

The inputs seem correct.
Even our code can run with PAW, but I am not sure whether the electron-phonon from QE can work properly with PAW or not.
So I suggest changing C.pbe-paw.upf to another NC or US for testing. In my example, I used C.pz-hgh.UPF.

[eamouj78]

Thank you for your reply.

Since the code did not complain, I thought it/QE works well with PAW. I will test with a NC PP and let you know.

Regards

[eamouj78]

Hi again,

I found what the problem is. The flags noinv and nosym must be set to true; otherwise the results are different. Something goes terribly wrong when the symmetries are switched on. I do not understand why. Is setting these flags to .true. a necessity for things to work?
Using the same C.pz-hgh.UPF PP:

Results with symmetry (nosym and noinv set to .false.)
#Raman spectra
#Laser energy 2.33 eV
#Raman shift (cm-1), Raman Spectra (XX,YY,ZZ,s+s+,s-s-,XY,YX,s+s-,s-s+)
1500.0000 0.10066241E+002 0.60146413E+003 0.12629235E+000 0.14578244E+003 0.14218476E+003 0.22148823E+002 0.27692059E+002 0.16844818E+003 0.20495587E+003
1500.1821 0.10146264E+002 0.60447220E+003 0.12725882E+000 0.14648761E+003 0.14285723E+003 0.22276515E+002 0.27865072E+002 0.16932385E+003 0.20609136E+003

Results with NO symmetry (nosym and noinv set to true.)
1500.0000 0.38484584E+003 0.38188234E+003 0.12586219E-003 0.12641939E-001 0.12509056E-001 0.38424858E+003 0.38426532E+003 0.76733604E+003 0.76788090E+003
1500.1821 0.38794558E+003 0.38495822E+003 0.12687503E-003 0.12743474E-001 0.12609523E-001 0.38734194E+003 0.38735882E+003 0.77351499E+003 0.77406423E+003

Regards

[eamouj78]

Hi again,

I just tried it with another PP, and I am getting the same results you reported.

The culprit (at least for the case of graphene) is indeed the nosym, noinv flags.

[nguyen-group]

Hi, I see. Sorry that I did not check scf.in carefully. No nosym and noinv must to set in the scf.in. The reason was explained in the paper. By using this option, both el-photon and el-phonon will be calculated with the same k- and q-mesh.

[eamouj78]

Yes, sorry, I have not seen that we need to set these flags to .true.

A final question: you mention that the flag "gamma" needs to be set according to the experiment.
Let us say there is no experimental reference, and the structure is still new theoretically; how can we choose values of "gamma" and "gamma_raman"? Do these need to be converged, or are the ones included in the examples OK?

Regards

[nguyen-group]

That is a good question. I don't think the convergence can work with gamma and gamma_raman. For a new material, a plot of Raman as a function of gamma might be helpful. You can see in Fig. of this paper: https://journals.aps.org/prb/supplemental/10.1103/PhysRevB.97.115407/tatsumi17_hrraman-SI.pdf

[eamouj78]

Thank you very much for your help and patience.
I will have a close look at that paper.

Regards