Hello<br><br>I have had a problem related to the Fermi level of a system of Silicon and Thallium. I was trying to compute a scf calculation of a Silicon slab made of 24 layers in the (1,1,1) direction with one thallium on top of it (the total length is about 45 Angstrom), with 10.41 Angstrom of vacuum in the same direction; this is my unit cell which tries to be a surface (fully relativistic noncollinear calculation with spin orbit coupling). First, I have used the choice<br>
<br> occupations='smearing',<br> smearing='methfessel-paxton',<br> degauss=0.05d0 <br><br>In this way I have let the system relax with the 'bfgs' method, and I have checked the convergence with respect to the cutoff energy (45 Ry) and the k_point mesh (16*16*1). As a result the system has ended up relaxed with negligible forces and I have calculated the bands using the results of the .save file, there was no problem when reading it and the bands fit well with the bulk background bands (calculated in bulk Si in 1 1 1 direction). Then I have increased the length of the vacuum to 30 Angstroms, I have relaxed the system, the atoms moved a little bit and I have checked all the convergences again. However, when I tried to use the results from the .save file to compute again the bands, the program has stopped with an error<br>
<br> from setup : error # 3<br> problem reading ef from file /lscratch/azpiroz/scf/scf.save<br><br>Also, when I have tried to perform a 'nscf' calculation using the results from the scf run, the program stops with<br>
<br> task # 5<br> from efermig : error # 1<br> internal error, cannot braket Ef<br>
<br>I have looked for possible explanations about this last error, and I have found that in the Users Guide from Espresso there is a place where it says<br><br><h2><a name="12868fc87570157d_SECTION000816000000000000000">
7.16 pw.x stops with "internal error: cannot braket Ef " in efermig</a>
</h2>
Possible reasons:
<ul><li>serious error in data, such as bad number of electrons,
insufficient number of bands, absurd value of broadening;
</li><li>the Fermi energy is found by bisection assuming that the
integrated DOS N(E ) is an increasing function of the energy. This
is not guaranteed for Methfessel-Paxton smearing of order 1 and can
give problems when very few k-points are used. Use some other
smearing function: simple Gaussian broadening or, better,
Marzari-Vanderbilt 'cold smearing'.
</li></ul>
<p>
I have tried many times the scf calculation and it never gives any error, the number of electrons is correct and I have used different values of the broadening and number of bands, and the error still persists in the nscf calculation. I dont have any clue about how to change 'correctly' the value of the broadening so that its not absurd, or the number of bands, any suggestion to this respect will be appreciated. Also, I have tried with Marzari-Vanderbilt smearing and many k points, but the error still persists. </p>
<p>The error seems to be related with the Fermi level and the type of occupation. The question is, why does it appear when the vacuum is 30 Angstrom but not when the vacuum is 10 Angstrom long? </p><p>Finally, I have seen that there is an option for the occupations called 'fixed', for 'insulators with a gap'. Even though this system is not an insulator (there is a band crossing the Fermi surface), I have tried this option and it works well, and even more it does not give any problem neither when computing the band structure nor with the 'nscf' calculation. The band structure obtained in this way coincides with the one obtained with 'methfessel-paxton' smearing (in the case of 10 Angstroms of vacuum), and as I said it works well for 30 Angstrom of vacuum. I think that the point is that with occupations='fixed' the program does not return any Fermi level, but I would really appreciate any help to this respect. Thank you very much in advance<br>
</p>Julen Ibaņez Azpiroz<br><br><br>University of the Basque Country<br>