help,how can I get the potential energy of each atom in a structure using DFT package

Message

weikexueerzhan · #1 Post by **weikexueerzhan** » Thu Mar 15, 2012 6:44 am

I am a freshman in DFT calculation. I don't understand the DFT method clearly,but I want to use DFT to do some calculations.

My question is that how how can I get the relative potential energy of each atom in a structure using DFT package ,which I have done by using classical MD method in Lammps(the command: compute 1 all pe/atom ).

In classical MD, the potential energy is calculated as the sum of pair, bond, angle, dihedral,improper, and kspace energy.).

We know that the energy of one atom in a structure will be different from others' for its particular coordinates. The energy of a atom will be affected by its nearby atoms(the background),so I want to know the relative potential energy of each atom in a structure.

I now know that i can get the force F of each atom from the results of vasp relaxation calculation .In the classical theory,F= dU/dr(U is the potential energy),,so if it's possible to get the potential energy from the VAsp calculation?

I want to know how can I do such a calculation.Please give some suggestions.

<span class='smallblacktext'>[ Edited ]</span>

b_lack · #2 Post by **b_lack** » Fri Mar 16, 2012 3:26 am

you just have to add
LVTOT = .TRUE.
on your INCAR file, and they will save your potential map in LOCPOT file

tlchan · #3 Post by **tlchan** » Fri Mar 16, 2012 4:37 am

For classical MD simulations, the energy is evaluated by summing two-body and three-body interaction potential functions between atoms. As such, the energy can be partitioned into contributions by different atoms. However, DFT calculations are performed differently. As far as I know, there is no standard method to partition the total energy into atomic contributions.

weikexueerzhan · #4 Post by **weikexueerzhan** » Mon Mar 19, 2012 3:44 am

[quote author= .TRUE.
on your INCAR file, and they will save your potential map in LOCPOT file[/quote]

B) Thanks for your reply. But is this total local potential comparable with the potential potential energy calculated in classical MD method,which is the sum of pair, bond, angle, dihedral,improper, and kspace energy?

weikexueerzhan · #5 Post by **weikexueerzhan** » Mon Mar 19, 2012 4:01 am

[quote="tlchan"]For classical MD simulations, the energy is evaluated by summing two-body and three-body interaction potential functions between atoms. As such, the energy can be partitioned into contributions by different atoms. However, DFT calculations are performed differently. As far as I know, there is no standard method to partition the total energy into atomic contributions. [/quote]
Thanks for your reply. This is also my consideration. How the DFT calculate the each atom energy is a problem. But I really need the calculation.

weikexueerzhan · #6 Post by **weikexueerzhan** » Thu Mar 22, 2012 10:25 am

[quote="b_lack"]you just have to add
LVTOT = .TRUE.
on your INCAR file, and they will save your potential map in LOCPOT file[/quote]

I have tested your method,but the LOCPOT gives not the potential energy but the work function.

b_lack · #7 Post by **b_lack** » Fri Mar 23, 2012 1:41 am

[quote="weikexueerzhan"][quote author= .TRUE.
on your INCAR file, and they will save your potential map in LOCPOT file[/quote]

This is not my method, this is command line

LVTOT --> L-V-TOT --> L(local) V(Potential) TOT(Total)
LOCPOT --> LOC-POT --> LOC(local) POT(potential)

You can read the detail on VASP manual p.45, section 5.19
Hope it will help

tlchan · #8 Post by **tlchan** » Fri Mar 23, 2012 4:20 am

It depends on what wei is really asking for. The LVTOT tag will request VASP to generate the LOCPOT output file. LOCPOT prints out the local part of the self-consistent potential as a function of position. There are various uses of the potential, including work function, band structure alignment, core-level shift etc.

From the first post of this thread, it seems like wei did some classical molecular dynamics simulations. Typically, they are performed using some empirical potentials that evaluate the interaction potential energy between atoms. I'm not sure if this potential energy is directly related to the self-consistent potential in DFT. The total potential energy of the whole system from a classical MD simulation should correspond to the total energy in DFT. Due to the way a classical simulation evaluates the energy, it turns out that the total potential energy can be decomposed easily into contributions from different atoms. On the other hand, DFT energies are evaluated very differently. In fact, we have difficulties telling which electron belongs to which atom, it just gets harder to assign part of the total energy to certain atoms.

As far as I know, there are somebody who tried to do what you want. Maybe you can do a literature search, or somebody else can offer their help.

b_lack · #9 Post by **b_lack** » Fri Mar 23, 2012 6:24 am

Hihihihi, ..After I googling around about "relative potential energy" I just realize that we have a different perspective about potential energy.

I'm still not to sure about this one, but "is it to see reaction path or conformations energy?", to see which structure is more stable or not based on energy level?

May be you can get those value from the OUTCAR file, search for the last TOTEN (Total Energy) and convert from eV to KJ/mol. Theres so many TOTEN on that file but the most stable energy is the last TOTEN data

You can also trace the energy change and compare them with the structure change from XDATCAR file

but this is still "maybe", I'm not too sure about Wei requirement

Hope it will help, ^_^

weikexueerzhan · #10 Post by **weikexueerzhan** » Tue Mar 27, 2012 4:48 am

[quote="weikexueerzhan"][quote="b_lack"]you just have to add
LVTOT = .TRUE.
on your INCAR file, and they will save your potential map in LOCPOT file[/quote]

This is not my method, this is command line

LVTOT --> L-V-TOT --> L(local) V(Potential) TOT(Total)
LOCPOT --> LOC-POT --> LOC(local) POT(potential)

You can read the detail on VASP manual p.45, section 5.19
Hope it will help[/quote]

Thanks. The local potential in vasp is not the energy of each atom.
how to decompose the total energy into the distribution of each atom may a theory problem of DFT .
It seems difficult.

weikexueerzhan · #11 Post by **weikexueerzhan** » Tue Mar 27, 2012 5:02 am

[quote="tlchan"]It depends on what wei is really asking for. The LVTOT tag will request VASP to generate the LOCPOT output file. LOCPOT prints out the local part of the self-consistent potential as a function of position. There are various uses of the potential, including work function, band structure alignment, core-level shift etc.

From the first post of this thread, it seems like wei did some classical molecular dynamics simulations. Typically, they are performed using some empirical potentials that evaluate the interaction potential energy between atoms. I'm not sure if this potential energy is directly related to the self-consistent potential in DFT. The total potential energy of the whole system from a classical MD simulation should correspond to the total energy in DFT. Due to the way a classical simulation evaluates the energy, it turns out that the total potential energy can be decomposed easily into contributions from different atoms. On the other hand, DFT energies are evaluated very differently. In fact, we have difficulties telling which electron belongs to which atom, it just gets harder to assign part of the total energy to certain atoms.

As far as I know, there are somebody who tried to do what you want. Maybe you can do a literature search, or somebody else can offer their help.[/quote]

Thanks. What you said is what I want. :|may it be not possible to decompose the energy into each atom distribution. May it need to be resolved by alter the DFT computation method.

weikexueerzhan · #12 Post by **weikexueerzhan** » Tue Mar 27, 2012 5:07 am

[quote="b_lack"]Hihihihi, ..After I googling around about "relative potential energy" I just realize that we have a different perspective about potential energy.

I'm still not to sure about this one, but "is it to see reaction path or conformations energy?", to see which structure is more stable or not based on energy level?

May be you can get those value from the OUTCAR file, search for the last TOTEN (Total Energy) and convert from eV to KJ/mol. Theres so many TOTEN on that file but the most stable energy is the last TOTEN data

You can also trace the energy change and compare them with the structure change from XDATCAR file

but this is still "maybe", I'm not too sure about Wei requirement

Hope it will help, ^_^[/quote]

haha O(∩_∩)O . thanks. what tlchan said is my requirement（my English is poor..⊙﹏⊙）. Vasp only gives the total energy of the system, I can,t find the each atom energy.