Significant Discrepancy Between ShengBTE and elphbolt (BTE-only) Results

[TiacWei]

I tested several structures (BN, BP, BAs, and BSb) using both ShengBTE and elphbolt (BTE only). However, I found that the thermal conductivity values are drastically different between the two codes.

For all calculations, I used the same espresso.ifc2 and FORCE_CONSTANTS_3RD files as input for both ShengBTE and elphbolt.

Here are the thermal conductivity (TC in W/mK) results:

ShengBTE:

• BAs: 193.06
• BN: 600.18
• BP: 475.99
• BSb: 110.95

elphbolt:

• BAs: 0.41
• BN: 282.06
• BP: 3.77
• BSb: 0.26

As a cross-check, I also tested the SiC example provided by both ShengBTE and elphbolt (again, BTE only), and the difference was about 100 W/mK, which is still quite significant.

Has anyone encountered similar discrepancies between ShengBTE and elphbolt results?

Any advice or suggestions would be greatly appreciated!

&allocations 
         nelements=2,
         natoms=2,
         ngrid(:)=20 20 20
&end 
&crystal 
         lfactor=0.1,
         lattvec(:,1)=0.000000000   2.403463720   2.403463720,
         lattvec(:,2)=2.403463720   0.000000000   2.403463720,
         lattvec(:,3)=2.403463720   2.403463720   0.000000000,
         elements='B' 'As',
         types=1  2,
         positions(:,1)=0.2500000000        0.2500000000        0.2500000000,
         positions(:,2)=0.0000000000        0.0000000000        0.0000000000,
         scell(:)=4 4 4
&end 
&parameters
        T=300
        scalebroad=1.0
        maxiter=50
&end
&flags
        isotopes=.TRUE.
        espresso=.true.
        nonanalytic=.FALSE.
        nanowires=.FALSE.
&end

&allocations
 numelements = 2
 numatoms = 2
/

&crystal_info
 name = 'BAs'
 elements ='B' 'As'
 atomtypes = 1 2
 VCA = .true.
 lattvecs(:,1) =  0.000000000   0.2403463720   0.2403463720
 lattvecs(:,2) =  0.2403463720   0.000000000   0.2403463720
 lattvecs(:,3) =  0.2403463720   0.2403463720   0.000000000
 basis(:,1) = 0.2500000000        0.2500000000        0.2500000000
 basis(:,2) = 0.0000000000        0.0000000000        0.0000000000
 T = 300.0 !K, crystal temperature
/

&numerics
 qmesh = 20 20 20
 onlyphbte = .t.
/

[nakib]

Dear TiacWei,

Thank you for your message.

The 3-phonon limited thermal conductivity of BAs is above 2000 W/m/K, which is 10x larger than what you got with your ShengBTE calculation. Did you confirm that your phonon dispersions look good? Are the lattice vectors and basis vectors consistent between your quantum espresso (harmonic and anharmonic) inputs and elphbolt inputs? Can you share a comparison of the phonon density of states calculated by quantum espresso and elphbolt? Can you share your elphbolt output?

By the way, for 3d materials, I suggest using the tetrahedron method in elphbolt.

Best,
Nakib