Dear all, I have a general question about calculating the transmission function for simple metals. On the example of ferromagnetic fcc cobalt, I performed a spin-polarized calculation of transmission using the latest version of PWCOND (v.5.1). Below are the input data for this system.
Input for the self-consistent calculation: &control calculation='scf', restart_mode='from_scratch', pseudo_dir = '/scratch/vborisov/pseudo/', outdir='/scratch/vborisov/tmp/Co-Transmission/', prefix='fct-2', wf_collect=.true. / &system ibrav = 6, celldm(1) = 7.35477531275, celldm(3) = 0.756973279, nat = 4, ntyp = 1, nspin = 2, nbnd = 40, starting_magnetization(1)=+1.80, ecutwfc = 63.0, ecutrho = 504.0, occupations='smearing', smearing='methfessel-paxton', degauss=0.02 / &electrons conv_thr = 1.0e-8 mixing_beta = 0.25 / ATOMIC_SPECIES Co 58.933 Co.pbe-nd-rrkjus.UPF ATOMIC_POSITIONS {crystal} Co 0.00 0.00 0.25 Co 0.50 0.50 0.25 Co 0.00 0.50 0.75 Co 0.50 0.00 0.75 K_POINTS {automatic} 15 15 20 0 0 0 One of the inputs for the transmission calculation: &inputcond outdir = '/scratch/vborisov/tmp/Co-Transmission', prefixl = 'fct-2', prefixs = 'fct-2', tran_file = 'TJ-k1550.Ef' ikind = 1, iofspin = 2, energy0 = 0.00d0, denergy = -0.01d0, ewind = 2.d0, epsproj = 1.d-5, delgep = 1.d-7, cutplot = 3.d0, nz1 = 22 / 1 0.00510204 0.13775510 1 1 At the end of the output file, one finds the following data: ngper, shell number = 271 271 ngper, n2d = 271 121 --- E-Ef = 0.0000000 k = 0.0051020 0.1377551 --- ie = 1 ik = 1 Nchannels of the left tip = 30 Right moving states: k1(2pi/a) k2(2pi/a) E-Ef (eV) -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 0.0000000 0.0000000 -0.0964682 -0.0000000 0.0000000 -0.0964689 -0.0000000 0.0000000 -0.0964705 0.0000000 0.0000000 -0.0964798 -0.0000000 0.0000000 -0.0965107 0.0000000 0.0000000 -0.0965482 0.0000000 0.0000000 -0.0966171 0.0000000 0.0000000 -0.0968639 0.0000000 0.0000000 -0.0985616 0.0000000 0.0000000 -0.1341194 0.0000003 0.0000000 -0.1855624 0.0000002 0.0000000 -0.2949790 0.0000002 0.0000000 -0.3297361 0.0000000 0.0000000 0.4517894 -0.0000001 0.0000000 Left moving states: k1(2pi/a) k2(2pi/a) E-Ef (eV) 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964682 -0.0000000 0.0000000 0.0964689 0.0000000 0.0000000 0.0964705 0.0000000 0.0000000 0.0964798 -0.0000000 0.0000000 0.0965107 0.0000000 0.0000000 0.0965482 0.0000000 0.0000000 0.0966171 0.0000000 0.0000000 0.0968639 0.0000000 0.0000000 0.0985616 0.0000000 0.0000000 0.1341194 0.0000003 0.0000000 0.1855624 0.0000002 0.0000000 0.2949790 0.0000002 0.0000000 0.3297361 0.0000000 0.0000000 -0.4517894 -0.0000001 0.0000000 to transmit Band j to band i transmissions and reflections: j i |T_ij|^2 |R_ij|^2 1 --> 1 1.00000 0.00000 1 --> 2 0.00000 0.00000 1 --> 3 0.00000 0.00000 1 --> 4 0.00000 0.00000 1 --> 5 0.00000 0.00000 1 --> 6 0.00000 0.00000 1 --> 7 0.00000 0.00000 1 --> 8 0.00000 0.00000 1 --> 9 0.00000 0.00000 1 --> 10 0.00000 0.00000 1 --> 11 0.00000 0.00000 1 --> 12 0.00000 0.00000 1 --> 13 0.00000 0.00000 1 --> 14 0.00000 0.00000 1 --> 15 0.00000 0.00000 1 --> 16 0.00000 0.00000 1 --> 17 0.00000 0.00000 1 --> 18 0.00000 0.00000 1 --> 19 0.00000 0.00000 1 --> 20 0.00000 0.00000 1 --> 21 0.00000 0.00000 1 --> 22 0.00000 0.00000 1 --> 23 0.00000 0.00000 1 --> 24 0.00000 0.00000 1 --> 25 0.00000 0.00000 1 --> 26 0.00000 0.00000 1 --> 27 0.00000 0.00000 1 --> 28 0.00000 0.00000 1 --> 29 0.00000 0.00000 1 --> 30 0.00000 0.00000 Total T_j, R_j = 1.00000 0.00000 2 --> 1 0.00000 0.00000 2 --> 2 1.00000 0.00000 2 --> 3 0.00000 0.00000 2 --> 4 0.00000 0.00000 2 --> 5 0.00000 0.00000 2 --> 6 0.00000 0.00000 2 --> 7 0.00000 0.00000 2 --> 8 0.00000 0.00000 2 --> 9 0.00000 0.00000 2 --> 10 0.00000 0.00000 2 --> 11 0.00000 0.00000 ... (the same for all other channels) 30 --> 24 0.00000 0.00000 30 --> 25 0.00000 0.00000 30 --> 26 0.00000 0.00000 30 --> 27 0.00000 0.00000 30 --> 28 0.00000 0.00000 30 --> 29 0.00000 0.00000 30 --> 30 1.00000 0.00000 Total T_j, R_j = 1.00000 0.00000 E-Ef(ev), T = 0.0000000 30.0000000 T_tot 0.00000 0.30000E+02 PWCOND : 1m25.79s CPU 3m 6.30s WALL init : 33.04s CPU 133.41s WALL ( 1 calls) poten : 0.02s CPU 0.02s WALL ( 2 calls) local : 2.43s CPU 2.46s WALL ( 1 calls) scatter_forw : 49.34s CPU 49.42s WALL ( 2 calls) compbs : 0.83s CPU 0.84s WALL ( 1 calls) compbs_2 : 0.62s CPU 0.63s WALL ( 1 calls) The transmission is way too large for this material, so the result is obviously wrong. However, if I set the epsproj parameter to a larger value, e.g. 10^-4, then I get the correct result T=4 (see below): ngper, shell number = 271 271 ngper, n2d = 271 69 --- E-Ef = 0.0000000 k = 0.0051020 0.1377551 --- ie = 1 ik = 1 Nchannels of the left tip = 4 Right moving states: k1(2pi/a) k2(2pi/a) E-Ef (eV) -0.1543659 0.0000003 0.0000000 -0.1880694 0.0000001 0.0000000 -0.2967059 0.0000002 0.0000000 0.4430621 -0.0000001 0.0000000 Left moving states: k1(2pi/a) k2(2pi/a) E-Ef (eV) 0.1543659 0.0000003 0.0000000 0.1880695 0.0000002 0.0000000 0.2967060 0.0000002 0.0000000 -0.4430621 -0.0000001 0.0000000 to transmit Band j to band i transmissions and reflections: j i |T_ij|^2 |R_ij|^2 1 --> 1 1.00000 0.00000 1 --> 2 0.00000 0.00000 1 --> 3 0.00000 0.00000 1 --> 4 0.00000 0.00000 Total T_j, R_j = 1.00000 0.00000 2 --> 1 0.00000 0.00000 2 --> 2 1.00000 0.00000 2 --> 3 0.00000 0.00000 2 --> 4 0.00000 0.00000 Total T_j, R_j = 1.00000 0.00000 3 --> 1 0.00000 0.00000 3 --> 2 0.00000 0.00000 3 --> 3 1.00000 0.00000 3 --> 4 0.00000 0.00000 Total T_j, R_j = 1.00000 0.00000 4 --> 1 0.00000 0.00000 4 --> 2 0.00000 0.00000 4 --> 3 0.00000 0.00000 4 --> 4 1.00000 0.00000 Total T_j, R_j = 1.00000 0.00000 E-Ef(ev), T = 0.0000000 4.0000000 T_tot 0.00000 0.40000E+01 PWCOND : 1m14.44s CPU 2m14.19s WALL init : 33.20s CPU 92.90s WALL ( 1 calls) poten : 0.02s CPU 0.02s WALL ( 2 calls) local : 1.86s CPU 1.88s WALL ( 1 calls) scatter_forw : 38.94s CPU 38.96s WALL ( 2 calls) compbs : 0.37s CPU 0.39s WALL ( 1 calls) compbs_2 : 0.29s CPU 0.30s WALL ( 1 calls) This situation is observed quite seldom (a few points in the BZ out of 1000). For example, for metals like Li there is no such problem at all. Also metal-insulator-metal systems in the tunneling regime do not show the aforementioned inconsistencies. I would be grateful, if somebody could give an explanation, as to how the epsproj parameter can influence the result of the calculation in such a dramatic way and how one can decide on the optimal value of this parameter. With kind regards, Vladislav Borisov Martin Luther University Halle-Wittenberg Von-Seckendorff-Platz 1, Room 1.17 06120, Halle (Saale), Germany Tel No: +49 (0) 345 55-25448 Fax No: +49 (0) 345 55-25446 Email: vladislav.borisov at physik.uni-halle.de