Discrepency between Molecular energy levels and Fermi energy and #of electron

[7we]

Hello,
   I am encountering a weird problem with ATK when I use the code (attached) to calculate the MOE, Fermi energy and # of electron and I would appreciate if someone can explain this. the problem is when I use this code several times, each time the result changes even though the code has not been changed.you can see this in the "C59W MOE.png" picture file.
the difference in the Fermi energy and MOE is really small so I might be able to ignore that (still I would love to know why), but the occupancy numbers are vastly different and can change the total properties of system.

And I would appreciate any suggestion that can enhance the accuracy of the calculation for this system (C59W) .

Cheers,
Ehsan

[zh]

If your molecule has HOMO-LUMO gap, you can use much smaller value   for "electron_temperature", i.e.,  which tends to  be fixed electron occupation.
 

[7we]

Thanks for the reply. I used 0.1 k as the temperature in my system, are you suggesting to use smaller value than 0.1 k?
And you didn't answer why I'm getting different result using the same code, would you please explain this too.

If your molecule has HOMO-LUMO gap, you can use much smaller value   for "electron_temperature", i.e.,  which tends to  be fixed electron occupation.
 

[zh]

The calculated system is C59W. The pseudoptentials for C and W you used are generated for the normal valence electron configurations, i.e., C: 2s^2  2p^2 ( 4 electrons); W: 6s^2 5d^4 (6 electrons). So the total number of valence electrons is 59*4 + 6 = 242. We expect the 121th molecular level (i.e., index of 120 in your attached picture) is the highest occupied one.  From your results, this level seems to be degenerated with the 122th molecular level ((i.e., index of 121 in your attached picture).  In such situation, which is contrast to what I said in my last reply, the use of "0.1 k" for electron temperature is not suitable. It would lead to a problem in determining the position of the highest occupied level. This is what you have met in your calculations.

Solution:
1) If you keep the use of non-spin polarization calculation (number_of_spins=1), you have to increase the value of electron temperature.
2) If you use the spin-polarization calculation(i.e., number_of_spins= 2), you can use small value of electron temperature as you already used.

The 2nd way is recommended because the spin-polarization calculation would give more reliable results.

[Anders Blom]

Thanks, zh, for noticing and clarifying!

[7we]

Thanks Zh, the second you solution that you mentioned solved my problem. Would you please explain a little bit more why do I need to use the spin polarized calculation? I tried to search for it but I couldn't find any answer.
 and do you think is there a better basis set that I can choose for my system? (I am currently using the Double Zeta Polarized)

Cheers

The calculated system is C59W. The pseudoptentials for C and W you used are generated for the normal valence electron configurations, i.e., C: 2s^2  2p^2 ( 4 electrons); W: 6s^2 5d^4 (6 electrons). So the total number of valence electrons is 59*4 + 6 = 242. We expect the 121th molecular level (i.e., index of 120 in your attached picture) is the highest occupied one.  From your results, this level seems to be degenerated with the 122th molecular level ((i.e., index of 121 in your attached picture).  In such situation, which is contrast to what I said in my last reply, the use of "0.1 k" for electron temperature is not suitable. It would lead to a problem in determining the position of the highest occupied level. This is what you have met in your calculations.

Solution:
1) If you keep the use of non-spin polarization calculation (number_of_spins=1), you have to increase the value of electron temperature.
2) If you use the spin-polarization calculation(i.e., number_of_spins= 2), you can use small value of electron temperature as you already used.

The 2nd way is recommended because the spin-polarization calculation would give more reliable results.

[zh]

The degeneracy of occupied level and unoccupied one can be easily broken by taking into account the spin exchange-correlation interaction, resulting in that the occupied level shifts toward lower energy and the occupied level shifts toward higher energy.

The highest occupied molecular level in your C59W system may be dominated mainly by the d orbital of W.  Your C59W is a open-shell configuration, where the spin-polarization calculations is preferred.
 
The use of DZP basis set is a safe choice.