Can CSLD get correct phonon dispersion of crystal with R3m phase

[davy12344]

Hi, Dr Zhou.
I read a series of your work about CSLD. I have tried to complete simulation on Si, NaCl and some crystals with tetragonal structure, and I got expected phonon dispersion, which matches to results from Phonopy. However, when I conducted on one material with R3m phase, I cannot get correct phonon dispersion. I am not sure that whether this software is confined to a particular crystalline phase such as cubic-like or tetragonal-like crystal or I made mistakes in some steps.

The following is my primitive cell where I used for calculated.
Primitive cell
1.0
2.5606386051259404 1.4783853879668094 2.8923474905900837
-2.5606386051259404 1.4783853879668094 2.8923474905900837
0.0000000000000000 -2.9567707759336188 2.8923474905900837
1 1
direct
0.8403047127269900 0.8403047127269900 0.8403047127269900
0.1596952632730222 0.1596952632730222 0.1596952632730223

To get cubic like supercell, I use following sc.txt
3 -3 0
3 3 -6
3 3 3

I use csld.in below:
prim = POSCAR
sym_tol = 1E-3
epsilon_inf = epsilon_inf.txt

born_charge = born_charge.txt
[model]
model_type = LD
cluster_in = clusters.out
cluster_out = clusters.out
symC_in = Cmat.mtx
symC_out = Cmat.mtx
max_order = 5
fractional_distance = False
cluster_diameter = 10 8 5 4

[training]
interface = VASP
corr_type = f
corr_in = Amat.mtx
corr_out = Amat.mtx
fval_in = fval.txt
fval_out = fval.txt
traindat1 = SPOSCAR dir_00000-disp0.01

[fitting]
solution_in = solution_all
solution_out = sol_2nd
nsubset = 5
holdsize = 0.09
method = 5

mulist = 1E-5 1E-7 1E-9 1E-11
maxIter = 300
tolerance = 1E-6
subsetsize = 0.85
lambda = 0.5
uscale_list = 0.01

submodel1 = harmonic 0 1 2
[phonon]
nac= 0

qpoint_fractional = False
wavevector = [[50,[ 0,0,0],'\Gamma',[ 0.5,0.5, 0.5],'Z', [0, 0.5, 0.5],'F', [0, 0, 0],'\Gamma', [0, -0.371921, 0.371921],'X', [0, 0, 0.5],'L']]
unit = meV

dos_grid = 15 15 15
nE_dos = 400
ismear= -1
epsilon = 0.05
pdos = True

thermal_T_range = 50 800 50
thermal_out = thermal_out.txt

[export_potential]
export_shengbte = 4 4 4 2 3 4

[prediction]
interface = VASP
corr_type = f
corr_in = Amat_pred.mtx
corr_out = Amat_pred.mtx
fval_in = fval_pred.txt
fval_out = fval_pred.txt
traindat0 = POSCAR traj*

##The result from CSLD

[feifzhou]

Please name the two elements in your POSCAR.

[davy12344]

Please name the two elements in your POSCAR.
Sorry. I add the name. The following is my primitive cell where I used for calculated.
Primitive cell
Ge1 Te1
1.0
2.5606386051259404 1.4783853879668094 2.8923474905900837
-2.5606386051259404 1.4783853879668094 2.8923474905900837
0.0000000000000000 -2.9567707759336188 2.8923474905900837
Ge Te
1 1
direct
0.8403047127269900 0.8403047127269900 0.8403047127269900 Ge
0.1596952632730222 0.1596952632730222 0.1596952632730223 Te